korg/linux
0
0
Fork 0
mirror of https://mirrors.bfsu.edu.cn/git/linux.git synced 2025-01-07 22:34:18 +08:00
Code Issues Packages Projects Releases Wiki Activity
7c8196fd43
linux/tools/power/x86/turbostat/turbostat.c

3848 lines
97 KiB
C
Raw Normal View History

tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
/*
* turbostat -- show CPU frequency and C-state residency
* on modern Intel turbo-capable processors.
*
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
* Copyright (c) 2013 Intel Corporation.
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
* Len Brown <len.brown@intel.com>
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License along with
* this program; if not, write to the Free Software Foundation, Inc.,
* 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
*/
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
#define _GNU_SOURCE
turbostat: Don't put unprocessed uapi headers in the include path turbostat's Makefile puts arch/x86/include/uapi/ in the include path, so that it can include <asm/msr.h> from it. It isn't in general safe to include even uapi headers directly from the kernel tree without processing them through scripts/headers_install.sh, but asm/msr.h happens to work. However, that include path can break with some versions of system headers, by overriding some system headers with the unprocessed versions directly from the kernel source. For instance: In file included from /build/x86-generic/usr/include/bits/sigcontext.h:28:0, from /build/x86-generic/usr/include/signal.h:339, from /build/x86-generic/usr/include/sys/wait.h:31, from turbostat.c:27: ../../../../arch/x86/include/uapi/asm/sigcontext.h:4:28: fatal error: linux/compiler.h: No such file or directory This occurs because the system bits/sigcontext.h on that build system includes <asm/sigcontext.h>, and asm/sigcontext.h in the kernel source includes <linux/compiler.h>, which scripts/headers_install.sh would have filtered out. Since turbostat really only wants a single header, just include that one header rather than putting an entire directory of kernel headers on the include path. In the process, switch from msr.h to msr-index.h, since turbostat just wants the MSR numbers. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Cc: stable@vger.kernel.org Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:12 +08:00
#include MSRHEADER
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
#include <stdarg.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
#include <stdio.h>
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
#include <err.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
#include <unistd.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <sys/stat.h>
#include <sys/resource.h>
#include <fcntl.h>
#include <signal.h>
#include <sys/time.h>
#include <stdlib.h>
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
#include <getopt.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
#include <dirent.h>
#include <string.h>
#include <ctype.h>
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
#include <sched.h>
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 11:44:48 +08:00
#include <time.h>
turbostat: Use GCC's CPUID functions to support PIC turbostat uses inline assembly to call cpuid. On 32-bit x86, on systems that have certain security features enabled by default that make -fPIC the default, this causes a build error: turbostat.c: In function ‘check_cpuid’: turbostat.c:1906:2: error: PIC register clobbered by ‘ebx’ in ‘asm’ asm("cpuid" : "=a" (fms), "=c" (ecx), "=d" (edx) : "a" (1) : "ebx"); ^ GCC provides a header cpuid.h, containing a __get_cpuid function that works with both PIC and non-PIC. (On PIC, it saves and restores ebx around the cpuid instruction.) Use that instead. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Cc: stable@vger.kernel.org Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:14 +08:00
#include <cpuid.h>
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
#include <linux/capability.h>
#include <errno.h>
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
char *proc_stat = "/proc/stat";
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
FILE *outf;
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
int *fd_percpu;
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 11:44:48 +08:00
struct timespec interval_ts = {5, 0};
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
unsigned int debug;
unsigned int rapl_joules;
unsigned int summary_only;
unsigned int dump_only;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
unsigned int skip_c0;
unsigned int skip_c1;
unsigned int do_nhm_cstates;
unsigned int do_snb_cstates;
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
unsigned int do_knl_cstates;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
unsigned int do_pc2;
unsigned int do_pc3;
unsigned int do_pc6;
unsigned int do_pc7;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
unsigned int do_c8_c9_c10;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
unsigned int do_skl_residency;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
unsigned int do_slm_cstates;
unsigned int use_c1_residency_msr;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
unsigned int has_aperf;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
unsigned int has_epb;
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:55 +08:00
unsigned int do_irtl_snb;
unsigned int do_irtl_hsw;
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
unsigned int units = 1000000; /* MHz etc */
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
unsigned int genuine_intel;
unsigned int has_invariant_tsc;
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
unsigned int do_nhm_platform_info;
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
unsigned int extra_msr_offset32;
unsigned int extra_msr_offset64;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
unsigned int extra_delta_offset32;
unsigned int extra_delta_offset64;
tools/power turbostat: KNL workaround for %Busy and Avg_MHz KNL increments APERF and MPERF every 1024 clocks. This is compliant with the architecture specification, which requires that only the ratio of APERF/MPERF need be valid. However, turbostat takes advantage of the fact that these two MSRs increment every un-halted clock at the actual and base frequency: AVG_MHz = APERF_delta/measurement_interval %Busy = MPERF_delta/TSC_delta This quirk is needed for these calculations to also work on KNL, which would otherwise show a value 1024x smaller than expected. Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-14 19:31:00 +08:00
unsigned int aperf_mperf_multiplier = 1;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
int do_irq = 1;
tools/power turbostat: display SMI count by default The SMI counter is popular -- so display it by default rather than requiring an option. What the heck, we've blown the 80 column budget on many systems already... Note that the value displayed is the delta during the measurement interval. The absolute value of the counter can still be seen with the generic 32-bit MSR option, ie. -m 0x34 Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-11 06:19:24 +08:00
int do_smi;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
double bclk;
tools/power turbostat: SKL: Adjust for TSC difference from base frequency On a Skylake with 1500MHz base frequency, the TSC runs at 1512MHz. This is because the TSC is no longer in the n*100 MHz BCLK domain, but is now in the m*24MHz crystal clock domain. (24 MHz * 63 = 1512 MHz) This adds error to several calculations in turbostat, unless the TSC sample sizes are adjusted for this difference. Note that calculations in the time domain are immune from this issue, as the timing sub-system has already calibrated the TSC against a known wall clock. AVG_MHz = APERF_delta/measurement_interval need no adjustment. APERF_delta is in the BCLK domain, and measurement_interval is in the time domain. TSC_MHz = TSC_delta/measurement_interval needs no adjustment -- as we really do want to report the actual measured TSC delta here, and measurement_interval is in the accurate time domain. %Busy = MPERF_delta/TSC_delta needs adjustment to use TSC_BCLK_DOMAIN_delta. TSC_BCLK_DOMAIN_delta = TSC_delta * base_hz / tsc_hz Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval need adjustment as above. No other metrics in turbostat need to be adjusted. Before: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 24.84 2216 1512 0 2191 98.73 2219 1514 2 0 0.01 2130 1512 1 9 0.43 2016 1512 3 2 0.08 2016 1512 After: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 25.05 2198 1512 0 2190 99.62 2199 1512 2 0 0.01 2152 1512 1 9 0.46 2000 1512 3 2 0.10 2000 1512 Note that in this example, the "Before" Bzy_MHz was reported as exceeding the 2200 max turbo rate. Also, even a pinned spin loop would not be reported as over 99% busy. Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-26 12:12:38 +08:00
double base_hz;
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-20 10:37:40 +08:00
unsigned int has_base_hz;
tools/power turbostat: SKL: Adjust for TSC difference from base frequency On a Skylake with 1500MHz base frequency, the TSC runs at 1512MHz. This is because the TSC is no longer in the n*100 MHz BCLK domain, but is now in the m*24MHz crystal clock domain. (24 MHz * 63 = 1512 MHz) This adds error to several calculations in turbostat, unless the TSC sample sizes are adjusted for this difference. Note that calculations in the time domain are immune from this issue, as the timing sub-system has already calibrated the TSC against a known wall clock. AVG_MHz = APERF_delta/measurement_interval need no adjustment. APERF_delta is in the BCLK domain, and measurement_interval is in the time domain. TSC_MHz = TSC_delta/measurement_interval needs no adjustment -- as we really do want to report the actual measured TSC delta here, and measurement_interval is in the accurate time domain. %Busy = MPERF_delta/TSC_delta needs adjustment to use TSC_BCLK_DOMAIN_delta. TSC_BCLK_DOMAIN_delta = TSC_delta * base_hz / tsc_hz Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval need adjustment as above. No other metrics in turbostat need to be adjusted. Before: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 24.84 2216 1512 0 2191 98.73 2219 1514 2 0 0.01 2130 1512 1 9 0.43 2016 1512 3 2 0.08 2016 1512 After: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 25.05 2198 1512 0 2190 99.62 2199 1512 2 0 0.01 2152 1512 1 9 0.46 2000 1512 3 2 0.10 2000 1512 Note that in this example, the "Before" Bzy_MHz was reported as exceeding the 2200 max turbo rate. Also, even a pinned spin loop would not be reported as over 99% busy. Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-26 12:12:38 +08:00
double tsc_tweak = 1.0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
unsigned int show_pkg;
unsigned int show_core;
unsigned int show_cpu;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
unsigned int show_pkg_only;
unsigned int show_core_only;
char *output_buffer, *outp;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
unsigned int do_rapl;
unsigned int do_dts;
unsigned int do_ptm;
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
unsigned int do_gfx_rc6_ms;
unsigned long long gfx_cur_rc6_ms;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
unsigned int do_gfx_mhz;
unsigned int gfx_cur_mhz;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
unsigned int tcc_activation_temp;
unsigned int tcc_activation_temp_override;
tools/power turbostat: correct DRAM RAPL units on recent Xeon processors While not yet documented in the Software Developer's Manual, the data-sheet for modern Xeon states that DRAM RAPL ENERGY units are fixed at 15.3 uJ, rather than being discovered via MSR. Before this patch, DRAM energy on these products is over-stated by turbostat because the RAPL units are 4x larger. ref: "Xeon E5-2600 v3/E5-1600 v3 Datasheet Volume 2" http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e5-v3-datasheet-vol-2.pdf Signed-off-by: Andrey Semin <andrey.semin@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-12-05 13:07:00 +08:00
double rapl_power_units, rapl_time_units;
double rapl_dram_energy_units, rapl_energy_units;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
double rapl_joule_counter_range;
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
unsigned int do_core_perf_limit_reasons;
unsigned int do_gfx_perf_limit_reasons;
unsigned int do_ring_perf_limit_reasons;
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
unsigned int crystal_hz;
unsigned long long tsc_hz;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
int base_cpu;
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-20 10:37:40 +08:00
double discover_bclk(unsigned int family, unsigned int model);
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
unsigned int has_hwp; /* IA32_PM_ENABLE, IA32_HWP_CAPABILITIES */
/* IA32_HWP_REQUEST, IA32_HWP_STATUS */
unsigned int has_hwp_notify; /* IA32_HWP_INTERRUPT */
unsigned int has_hwp_activity_window; /* IA32_HWP_REQUEST[bits 41:32] */
unsigned int has_hwp_epp; /* IA32_HWP_REQUEST[bits 31:24] */
unsigned int has_hwp_pkg; /* IA32_HWP_REQUEST_PKG */
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
turbostat: run on HSX Haswell Xeon has slightly different RAPL support than client HSW, which prevented the previous version of turbostat from running on HSX. Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-03 15:19:19 +08:00
#define RAPL_PKG (1 << 0)
/* 0x610 MSR_PKG_POWER_LIMIT */
/* 0x611 MSR_PKG_ENERGY_STATUS */
#define RAPL_PKG_PERF_STATUS (1 << 1)
/* 0x613 MSR_PKG_PERF_STATUS */
#define RAPL_PKG_POWER_INFO (1 << 2)
/* 0x614 MSR_PKG_POWER_INFO */
#define RAPL_DRAM (1 << 3)
/* 0x618 MSR_DRAM_POWER_LIMIT */
/* 0x619 MSR_DRAM_ENERGY_STATUS */
#define RAPL_DRAM_PERF_STATUS (1 << 4)
/* 0x61b MSR_DRAM_PERF_STATUS */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
#define RAPL_DRAM_POWER_INFO (1 << 5)
/* 0x61c MSR_DRAM_POWER_INFO */
turbostat: run on HSX Haswell Xeon has slightly different RAPL support than client HSW, which prevented the previous version of turbostat from running on HSX. Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-03 15:19:19 +08:00
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
#define RAPL_CORES (1 << 6)
turbostat: run on HSX Haswell Xeon has slightly different RAPL support than client HSW, which prevented the previous version of turbostat from running on HSX. Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-03 15:19:19 +08:00
/* 0x638 MSR_PP0_POWER_LIMIT */
/* 0x639 MSR_PP0_ENERGY_STATUS */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
#define RAPL_CORE_POLICY (1 << 7)
turbostat: run on HSX Haswell Xeon has slightly different RAPL support than client HSW, which prevented the previous version of turbostat from running on HSX. Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-03 15:19:19 +08:00
/* 0x63a MSR_PP0_POLICY */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
#define RAPL_GFX (1 << 8)
turbostat: run on HSX Haswell Xeon has slightly different RAPL support than client HSW, which prevented the previous version of turbostat from running on HSX. Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-03 15:19:19 +08:00
/* 0x640 MSR_PP1_POWER_LIMIT */
/* 0x641 MSR_PP1_ENERGY_STATUS */
/* 0x642 MSR_PP1_POLICY */
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
#define TJMAX_DEFAULT 100
#define MAX(a, b) ((a) > (b) ? (a) : (b))
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
int aperf_mperf_unstable;
int backwards_count;
char *progname;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
cpu_set_t *cpu_present_set, *cpu_affinity_set;
size_t cpu_present_setsize, cpu_affinity_setsize;
struct thread_data {
unsigned long long tsc;
unsigned long long aperf;
unsigned long long mperf;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
unsigned long long c1;
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
unsigned long long extra_msr64;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
unsigned long long extra_delta64;
unsigned long long extra_msr32;
unsigned long long extra_delta32;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
unsigned int irq_count;
tools/power turbostat: display SMI count by default The SMI counter is popular -- so display it by default rather than requiring an option. What the heck, we've blown the 80 column budget on many systems already... Note that the value displayed is the delta during the measurement interval. The absolute value of the counter can still be seen with the generic 32-bit MSR option, ie. -m 0x34 Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-11 06:19:24 +08:00
unsigned int smi_count;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
unsigned int cpu_id;
unsigned int flags;
#define CPU_IS_FIRST_THREAD_IN_CORE 0x2
#define CPU_IS_FIRST_CORE_IN_PACKAGE 0x4
} *thread_even, *thread_odd;
struct core_data {
unsigned long long c3;
unsigned long long c6;
unsigned long long c7;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
unsigned int core_temp_c;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
unsigned int core_id;
} *core_even, *core_odd;
struct pkg_data {
unsigned long long pc2;
unsigned long long pc3;
unsigned long long pc6;
unsigned long long pc7;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
unsigned long long pc8;
unsigned long long pc9;
unsigned long long pc10;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
unsigned long long pkg_wtd_core_c0;
unsigned long long pkg_any_core_c0;
unsigned long long pkg_any_gfxe_c0;
unsigned long long pkg_both_core_gfxe_c0;
tools/power turbostat: work around RC6 counter wrap Sometimes the rc6 sysfs counter spontaneously resets, causing turbostat prints a very large number as it tries to calcuate % = 100 * (old - new) / interval When we see (old > new), print ***.**% instead of a bogus huge number. Note that this detection is not fool-proof, as the counter could reset several times and still result in new > old. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:16:00 +08:00
long long gfx_rc6_ms;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
unsigned int gfx_mhz;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
unsigned int package_id;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
unsigned int energy_pkg; /* MSR_PKG_ENERGY_STATUS */
unsigned int energy_dram; /* MSR_DRAM_ENERGY_STATUS */
unsigned int energy_cores; /* MSR_PP0_ENERGY_STATUS */
unsigned int energy_gfx; /* MSR_PP1_ENERGY_STATUS */
unsigned int rapl_pkg_perf_status; /* MSR_PKG_PERF_STATUS */
unsigned int rapl_dram_perf_status; /* MSR_DRAM_PERF_STATUS */
unsigned int pkg_temp_c;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
} *package_even, *package_odd;
#define ODD_COUNTERS thread_odd, core_odd, package_odd
#define EVEN_COUNTERS thread_even, core_even, package_even
#define GET_THREAD(thread_base, thread_no, core_no, pkg_no) \
(thread_base + (pkg_no) * topo.num_cores_per_pkg * \
topo.num_threads_per_core + \
(core_no) * topo.num_threads_per_core + (thread_no))
#define GET_CORE(core_base, core_no, pkg_no) \
(core_base + (pkg_no) * topo.num_cores_per_pkg + (core_no))
#define GET_PKG(pkg_base, pkg_no) (pkg_base + pkg_no)
struct system_summary {
struct thread_data threads;
struct core_data cores;
struct pkg_data packages;
} sum, average;
struct topo_params {
int num_packages;
int num_cpus;
int num_cores;
int max_cpu_num;
int num_cores_per_pkg;
int num_threads_per_core;
} topo;
struct timeval tv_even, tv_odd, tv_delta;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
int *irq_column_2_cpu; /* /proc/interrupts column numbers */
int *irqs_per_cpu; /* indexed by cpu_num */
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
void setup_all_buffers(void);
int cpu_is_not_present(int cpu)
tools/power turbostat: fix un-intended affinity of forked program Linux 3.4 included a modification to turbostat to lower cross-call overhead by using scheduler affinity: 15aaa34654831e98dd76f7738b6c7f5d05a66430 (tools turbostat: reduce measurement overhead due to IPIs) In the use-case where turbostat forks a child program, that change had the un-intended side-effect of binding the child to the last cpu in the system. This change removed the binding before forking the child. This is a back-port of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 11:24:00 +08:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return !CPU_ISSET_S(cpu, cpu_present_setsize, cpu_present_set);
tools/power turbostat: fix un-intended affinity of forked program Linux 3.4 included a modification to turbostat to lower cross-call overhead by using scheduler affinity: 15aaa34654831e98dd76f7738b6c7f5d05a66430 (tools turbostat: reduce measurement overhead due to IPIs) In the use-case where turbostat forks a child program, that change had the un-intended side-effect of binding the child to the last cpu in the system. This change removed the binding before forking the child. This is a back-port of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 11:24:00 +08:00
}
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
/*
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
* run func(thread, core, package) in topology order
* skip non-present cpus
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
*/
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int for_all_cpus(int (func)(struct thread_data *, struct core_data *, struct pkg_data *),
struct thread_data *thread_base, struct core_data *core_base, struct pkg_data *pkg_base)
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int retval, pkg_no, core_no, thread_no;
tools/power turbostat: fix un-intended affinity of forked program Linux 3.4 included a modification to turbostat to lower cross-call overhead by using scheduler affinity: 15aaa34654831e98dd76f7738b6c7f5d05a66430 (tools turbostat: reduce measurement overhead due to IPIs) In the use-case where turbostat forks a child program, that change had the un-intended side-effect of binding the child to the last cpu in the system. This change removed the binding before forking the child. This is a back-port of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 11:24:00 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
for (pkg_no = 0; pkg_no < topo.num_packages; ++pkg_no) {
for (core_no = 0; core_no < topo.num_cores_per_pkg; ++core_no) {
for (thread_no = 0; thread_no <
topo.num_threads_per_core; ++thread_no) {
struct thread_data *t;
struct core_data *c;
struct pkg_data *p;
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
t = GET_THREAD(thread_base, thread_no, core_no, pkg_no);
if (cpu_is_not_present(t->cpu_id))
continue;
c = GET_CORE(core_base, core_no, pkg_no);
p = GET_PKG(pkg_base, pkg_no);
retval = func(t, c, p);
if (retval)
return retval;
}
}
}
return 0;
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
}
int cpu_migrate(int cpu)
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
CPU_ZERO_S(cpu_affinity_setsize, cpu_affinity_set);
CPU_SET_S(cpu, cpu_affinity_setsize, cpu_affinity_set);
if (sched_setaffinity(0, cpu_affinity_setsize, cpu_affinity_set) == -1)
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
return -1;
else
return 0;
}
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
int get_msr_fd(int cpu)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
char pathname[32];
int fd;
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
fd = fd_percpu[cpu];
if (fd)
return fd;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
sprintf(pathname, "/dev/cpu/%d/msr", cpu);
fd = open(pathname, O_RDONLY);
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
if (fd < 0)
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
err(-1, "%s open failed, try chown or chmod +r /dev/cpu/*/msr, or run as root", pathname);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
fd_percpu[cpu] = fd;
return fd;
}
int get_msr(int cpu, off_t offset, unsigned long long *msr)
{
ssize_t retval;
retval = pread(get_msr_fd(cpu), msr, sizeof(*msr), offset);
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
if (retval != sizeof *msr)
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
err(-1, "msr %d offset 0x%llx read failed", cpu, (unsigned long long)offset);
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
/*
* Example Format w/ field column widths:
*
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
* Package Core CPU Avg_MHz Bzy_MHz TSC_MHz IRQ SMI Busy% CPU_%c1 CPU_%c3 CPU_%c6 CPU_%c7 CoreTmp PkgTmp GFXMHz Pkg%pc2 Pkg%pc3 Pkg%pc6 Pkg%pc7 PkgWatt CorWatt GFXWatt
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
* 12345678123456781234567812345678123456781234567812345678123456781234567812345678123456781234567812345678123456781234567812345678123456781234567812345678123456781234567812345678
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
*/
tools: turbostat: style updates Follow kernel coding style traditions more closely. Delete typedef, re-name "per cpu counters" to simply be counters etc. This patch changes no functionality. Suggested-by: Thiago Farina <tfransosi@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
2011-02-11 12:36:34 +08:00
void print_header(void)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
if (show_pkg)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Package");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (show_core)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Core");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (show_cpu)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " CPU");
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
if (has_aperf)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Avg_MHz");
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
if (has_aperf)
tools/power turbostat: re-name "%Busy" field to "Busy%" some tools processing turbostat output have difficulty with items that begin with %... Reported-by: Jacob Pan <jacob.jun.pan@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:41:53 +08:00
outp += sprintf(outp, " Busy%%");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (has_aperf)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Bzy_MHz");
outp += sprintf(outp, " TSC_MHz");
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 13:34:57 +08:00
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
if (extra_delta_offset32)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " count 0x%03X", extra_delta_offset32);
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
if (extra_delta_offset64)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " COUNT 0x%03X", extra_delta_offset64);
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
if (extra_msr_offset32)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " MSR 0x%03X", extra_msr_offset32);
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
if (extra_msr_offset64)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " MSR 0x%03X", extra_msr_offset64);
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 13:34:57 +08:00
if (!debug)
goto done;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
if (do_irq)
outp += sprintf(outp, " IRQ");
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 13:34:57 +08:00
if (do_smi)
outp += sprintf(outp, " SMI");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (do_nhm_cstates)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " CPU%%c1");
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
if (do_nhm_cstates && !do_slm_cstates && !do_knl_cstates)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " CPU%%c3");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (do_nhm_cstates)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " CPU%%c6");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (do_snb_cstates)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " CPU%%c7");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (do_dts)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " CoreTmp");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (do_ptm)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " PkgTmp");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
if (do_gfx_rc6_ms)
outp += sprintf(outp, " GFX%%rc6");
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
if (do_gfx_mhz)
outp += sprintf(outp, " GFXMHz");
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
if (do_skl_residency) {
outp += sprintf(outp, " Totl%%C0");
outp += sprintf(outp, " Any%%C0");
outp += sprintf(outp, " GFX%%C0");
outp += sprintf(outp, " CPUGFX%%");
}
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc2)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Pkg%%pc2");
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc3)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Pkg%%pc3");
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc6)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Pkg%%pc6");
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc7)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Pkg%%pc7");
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
if (do_c8_c9_c10) {
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Pkg%%pc8");
outp += sprintf(outp, " Pkg%%pc9");
outp += sprintf(outp, " Pk%%pc10");
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl && !rapl_joules) {
if (do_rapl & RAPL_PKG)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " PkgWatt");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_CORES)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " CorWatt");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_GFX)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " GFXWatt");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_DRAM)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " RAMWatt");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_PKG_PERF_STATUS)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " PKG_%%");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_DRAM_PERF_STATUS)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " RAM_%%");
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
} else if (do_rapl && rapl_joules) {
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_PKG)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Pkg_J");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_CORES)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " Cor_J");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_GFX)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " GFX_J");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_DRAM)
tools/power turbostat: fix typo on DRAM column in Joules-mode < RAM_W > RAM_J Reported-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-07-24 22:35:23 +08:00
outp += sprintf(outp, " RAM_J");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_PKG_PERF_STATUS)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " PKG_%%");
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_DRAM_PERF_STATUS)
tools/power turbostat: tweak whitespace in output format turbostat -S output was off by 1 space before this patch. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 09:22:13 +08:00
outp += sprintf(outp, " RAM_%%");
outp += sprintf(outp, " time");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
}
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 13:34:57 +08:00
done:
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
outp += sprintf(outp, "\n");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int dump_counters(struct thread_data *t, struct core_data *c,
struct pkg_data *p)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "t %p, c %p, p %p\n", t, c, p);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (t) {
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "CPU: %d flags 0x%x\n",
t->cpu_id, t->flags);
outp += sprintf(outp, "TSC: %016llX\n", t->tsc);
outp += sprintf(outp, "aperf: %016llX\n", t->aperf);
outp += sprintf(outp, "mperf: %016llX\n", t->mperf);
outp += sprintf(outp, "c1: %016llX\n", t->c1);
outp += sprintf(outp, "msr0x%x: %08llX\n",
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
extra_delta_offset32, t->extra_delta32);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "msr0x%x: %016llX\n",
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
extra_delta_offset64, t->extra_delta64);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "msr0x%x: %08llX\n",
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
extra_msr_offset32, t->extra_msr32);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "msr0x%x: %016llX\n",
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
extra_msr_offset64, t->extra_msr64);
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
if (do_irq)
outp += sprintf(outp, "IRQ: %08X\n", t->irq_count);
tools/power turbostat: display SMI count by default The SMI counter is popular -- so display it by default rather than requiring an option. What the heck, we've blown the 80 column budget on many systems already... Note that the value displayed is the delta during the measurement interval. The absolute value of the counter can still be seen with the generic 32-bit MSR option, ie. -m 0x34 Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-11 06:19:24 +08:00
if (do_smi)
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "SMI: %08X\n", t->smi_count);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (c) {
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "core: %d\n", c->core_id);
outp += sprintf(outp, "c3: %016llX\n", c->c3);
outp += sprintf(outp, "c6: %016llX\n", c->c6);
outp += sprintf(outp, "c7: %016llX\n", c->c7);
outp += sprintf(outp, "DTS: %dC\n", c->core_temp_c);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (p) {
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "package: %d\n", p->package_id);
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
outp += sprintf(outp, "Weighted cores: %016llX\n", p->pkg_wtd_core_c0);
outp += sprintf(outp, "Any cores: %016llX\n", p->pkg_any_core_c0);
outp += sprintf(outp, "Any GFX: %016llX\n", p->pkg_any_gfxe_c0);
outp += sprintf(outp, "CPU + GFX: %016llX\n", p->pkg_both_core_gfxe_c0);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "pc2: %016llX\n", p->pc2);
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc3)
outp += sprintf(outp, "pc3: %016llX\n", p->pc3);
if (do_pc6)
outp += sprintf(outp, "pc6: %016llX\n", p->pc6);
if (do_pc7)
outp += sprintf(outp, "pc7: %016llX\n", p->pc7);
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "pc8: %016llX\n", p->pc8);
outp += sprintf(outp, "pc9: %016llX\n", p->pc9);
outp += sprintf(outp, "pc10: %016llX\n", p->pc10);
outp += sprintf(outp, "Joules PKG: %0X\n", p->energy_pkg);
outp += sprintf(outp, "Joules COR: %0X\n", p->energy_cores);
outp += sprintf(outp, "Joules GFX: %0X\n", p->energy_gfx);
outp += sprintf(outp, "Joules RAM: %0X\n", p->energy_dram);
outp += sprintf(outp, "Throttle PKG: %0X\n",
p->rapl_pkg_perf_status);
outp += sprintf(outp, "Throttle RAM: %0X\n",
p->rapl_dram_perf_status);
outp += sprintf(outp, "PTM: %dC\n", p->pkg_temp_c);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
outp += sprintf(outp, "\n");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools turbostat: add summary option turbostat -s cuts down on the amount of output, per user request. also treak some output whitespace and the man page. Signed-off-by: Len Brown <len.brown@intel.com>
2012-02-07 07:37:16 +08:00
/*
* column formatting convention & formats
*/
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int format_counters(struct thread_data *t, struct core_data *c,
struct pkg_data *p)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
double interval_float;
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
char *fmt8;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* if showing only 1st thread in core and this isn't one, bail out */
if (show_core_only && !(t->flags & CPU_IS_FIRST_THREAD_IN_CORE))
return 0;
/* if showing only 1st thread in pkg and this isn't one, bail out */
if (show_pkg_only && !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
interval_float = tv_delta.tv_sec + tv_delta.tv_usec/1000000.0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* topo columns, print blanks on 1st (average) line */
if (t == &average.threads) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (show_pkg)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, " -");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (show_core)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, " -");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (show_cpu)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, " -");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
} else {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (show_pkg) {
if (p)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8d", p->package_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
else
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, " -");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
if (show_core) {
if (c)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8d", c->core_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
else
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, " -");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (show_cpu)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8d", t->cpu_id);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
/* Avg_MHz */
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
if (has_aperf)
outp += sprintf(outp, "%8.0f",
1.0 / units * t->aperf / interval_float);
tools/power turbostat: re-name "%Busy" field to "Busy%" some tools processing turbostat output have difficulty with items that begin with %... Reported-by: Jacob Pan <jacob.jun.pan@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:41:53 +08:00
/* Busy% */
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
if (has_aperf) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (!skip_c0)
tools/power turbostat: SKL: Adjust for TSC difference from base frequency On a Skylake with 1500MHz base frequency, the TSC runs at 1512MHz. This is because the TSC is no longer in the n*100 MHz BCLK domain, but is now in the m*24MHz crystal clock domain. (24 MHz * 63 = 1512 MHz) This adds error to several calculations in turbostat, unless the TSC sample sizes are adjusted for this difference. Note that calculations in the time domain are immune from this issue, as the timing sub-system has already calibrated the TSC against a known wall clock. AVG_MHz = APERF_delta/measurement_interval need no adjustment. APERF_delta is in the BCLK domain, and measurement_interval is in the time domain. TSC_MHz = TSC_delta/measurement_interval needs no adjustment -- as we really do want to report the actual measured TSC delta here, and measurement_interval is in the accurate time domain. %Busy = MPERF_delta/TSC_delta needs adjustment to use TSC_BCLK_DOMAIN_delta. TSC_BCLK_DOMAIN_delta = TSC_delta * base_hz / tsc_hz Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval need adjustment as above. No other metrics in turbostat need to be adjusted. Before: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 24.84 2216 1512 0 2191 98.73 2219 1514 2 0 0.01 2130 1512 1 9 0.43 2016 1512 3 2 0.08 2016 1512 After: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 25.05 2198 1512 0 2190 99.62 2199 1512 2 0 0.01 2152 1512 1 9 0.46 2000 1512 3 2 0.10 2000 1512 Note that in this example, the "Before" Bzy_MHz was reported as exceeding the 2200 max turbo rate. Also, even a pinned spin loop would not be reported as over 99% busy. Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-26 12:12:38 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * t->mperf/t->tsc/tsc_tweak);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
else
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "********");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
/* Bzy_MHz */
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-20 10:37:40 +08:00
if (has_aperf) {
if (has_base_hz)
outp += sprintf(outp, "%8.0f", base_hz / units * t->aperf / t->mperf);
else
outp += sprintf(outp, "%8.0f",
1.0 * t->tsc / units * t->aperf / t->mperf / interval_float);
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
/* TSC_MHz */
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.0f", 1.0 * t->tsc/units/interval_float);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
/* delta */
if (extra_delta_offset32)
outp += sprintf(outp, " %11llu", t->extra_delta32);
/* DELTA */
if (extra_delta_offset64)
outp += sprintf(outp, " %11llu", t->extra_delta64);
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
/* msr */
if (extra_msr_offset32)
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
outp += sprintf(outp, " 0x%08llx", t->extra_msr32);
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
tools/power turbostat: make -M output pretty The -M option dumps the specified 64-bit MSR with every sample. Previously it was output at the end of each line. However, with the v2 style of printing, the lines are now staggered, making MSR output hard to read. So move the MSR output column to the left where things are aligned. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 10:56:06 +08:00
/* MSR */
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
if (extra_msr_offset64)
outp += sprintf(outp, " 0x%016llx", t->extra_msr64);
tools/power turbostat: make -M output pretty The -M option dumps the specified 64-bit MSR with every sample. Previously it was output at the end of each line. However, with the v2 style of printing, the lines are now staggered, making MSR output hard to read. So move the MSR output column to the left where things are aligned. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 10:56:06 +08:00
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 13:34:57 +08:00
if (!debug)
goto done;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
/* IRQ */
if (do_irq)
outp += sprintf(outp, "%8d", t->irq_count);
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 13:34:57 +08:00
/* SMI */
if (do_smi)
outp += sprintf(outp, "%8d", t->smi_count);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (do_nhm_cstates) {
if (!skip_c1)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * t->c1/t->tsc);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
else
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "********");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* print per-core data only for 1st thread in core */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE))
goto done;
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
if (do_nhm_cstates && !do_slm_cstates && !do_knl_cstates)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * c->c3/t->tsc);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (do_nhm_cstates)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * c->c6/t->tsc);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (do_snb_cstates)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * c->c7/t->tsc);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (do_dts)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8d", c->core_temp_c);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* print per-package data only for 1st core in package */
if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
goto done;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
/* PkgTmp */
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (do_ptm)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8d", p->pkg_temp_c);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
/* GFXrc6 */
tools/power turbostat: work around RC6 counter wrap Sometimes the rc6 sysfs counter spontaneously resets, causing turbostat prints a very large number as it tries to calcuate % = 100 * (old - new) / interval When we see (old > new), print ***.**% instead of a bogus huge number. Note that this detection is not fool-proof, as the counter could reset several times and still result in new > old. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:16:00 +08:00
if (do_gfx_rc6_ms) {
if (p->gfx_rc6_ms == -1) { /* detect counter reset */
outp += sprintf(outp, " ***.**");
} else {
outp += sprintf(outp, "%8.2f",
p->gfx_rc6_ms / 10.0 / interval_float);
}
}
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
/* GFXMHz */
if (do_gfx_mhz)
outp += sprintf(outp, "%8d", p->gfx_mhz);
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
/* Totl%C0, Any%C0 GFX%C0 CPUGFX% */
if (do_skl_residency) {
outp += sprintf(outp, "%8.2f", 100.0 * p->pkg_wtd_core_c0/t->tsc);
outp += sprintf(outp, "%8.2f", 100.0 * p->pkg_any_core_c0/t->tsc);
outp += sprintf(outp, "%8.2f", 100.0 * p->pkg_any_gfxe_c0/t->tsc);
outp += sprintf(outp, "%8.2f", 100.0 * p->pkg_both_core_gfxe_c0/t->tsc);
}
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc2)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * p->pc2/t->tsc);
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc3)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * p->pc3/t->tsc);
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc6)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * p->pc6/t->tsc);
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc7)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * p->pc7/t->tsc);
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
if (do_c8_c9_c10) {
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, "%8.2f", 100.0 * p->pc8/t->tsc);
outp += sprintf(outp, "%8.2f", 100.0 * p->pc9/t->tsc);
outp += sprintf(outp, "%8.2f", 100.0 * p->pc10/t->tsc);
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
/*
* If measurement interval exceeds minimum RAPL Joule Counter range,
* indicate that results are suspect by printing "**" in fraction place.
*/
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
if (interval_float < rapl_joule_counter_range)
fmt8 = "%8.2f";
else
fmt8 = " %6.0f**";
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl && !rapl_joules) {
if (do_rapl & RAPL_PKG)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8, p->energy_pkg * rapl_energy_units / interval_float);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_CORES)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8, p->energy_cores * rapl_energy_units / interval_float);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_GFX)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8, p->energy_gfx * rapl_energy_units / interval_float);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_DRAM)
tools/power turbostat: correct DRAM RAPL units on recent Xeon processors While not yet documented in the Software Developer's Manual, the data-sheet for modern Xeon states that DRAM RAPL ENERGY units are fixed at 15.3 uJ, rather than being discovered via MSR. Before this patch, DRAM energy on these products is over-stated by turbostat because the RAPL units are 4x larger. ref: "Xeon E5-2600 v3/E5-1600 v3 Datasheet Volume 2" http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e5-v3-datasheet-vol-2.pdf Signed-off-by: Andrey Semin <andrey.semin@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-12-05 13:07:00 +08:00
outp += sprintf(outp, fmt8, p->energy_dram * rapl_dram_energy_units / interval_float);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_PKG_PERF_STATUS)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8, 100.0 * p->rapl_pkg_perf_status * rapl_time_units / interval_float);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_DRAM_PERF_STATUS)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8, 100.0 * p->rapl_dram_perf_status * rapl_time_units / interval_float);
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
} else if (do_rapl && rapl_joules) {
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_PKG)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8,
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
p->energy_pkg * rapl_energy_units);
if (do_rapl & RAPL_CORES)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8,
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
p->energy_cores * rapl_energy_units);
if (do_rapl & RAPL_GFX)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8,
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
p->energy_gfx * rapl_energy_units);
if (do_rapl & RAPL_DRAM)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8,
tools/power turbostat: correct DRAM RAPL units on recent Xeon processors While not yet documented in the Software Developer's Manual, the data-sheet for modern Xeon states that DRAM RAPL ENERGY units are fixed at 15.3 uJ, rather than being discovered via MSR. Before this patch, DRAM energy on these products is over-stated by turbostat because the RAPL units are 4x larger. ref: "Xeon E5-2600 v3/E5-1600 v3 Datasheet Volume 2" http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e5-v3-datasheet-vol-2.pdf Signed-off-by: Andrey Semin <andrey.semin@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-12-05 13:07:00 +08:00
p->energy_dram * rapl_dram_energy_units);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_PKG_PERF_STATUS)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8, 100.0 * p->rapl_pkg_perf_status * rapl_time_units / interval_float);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
if (do_rapl & RAPL_DRAM_PERF_STATUS)
tools/power turbostat: simplify output, add Avg_MHz Use 8 columns for each number ouput. We don't fit into 80 columns on most machines, so keep the format simple. Print frequency in MHz instead of GHz. We've got 8 columns now, so use them to show low frequency in a more natural unit. Many users didn't understand what %c0 meant, so re-name it to be %Busy. Add Avg_MHz column, which is the frequency that many users expect to see -- the total number of cycles executed over the measurement interval. People found the previous GHz to be confusing, since it was the speed only over the non-idle interval. That measurement has been re-named Bzy_MHz. Suggested-by: Dirk J. Brandewie Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-06 13:55:19 +08:00
outp += sprintf(outp, fmt8, 100.0 * p->rapl_dram_perf_status * rapl_time_units / interval_float);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
outp += sprintf(outp, fmt8, interval_float);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
done:
outp += sprintf(outp, "\n");
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
void flush_output_stdout(void)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
{
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
FILE *filep;
if (outf == stderr)
filep = stdout;
else
filep = outf;
fputs(output_buffer, filep);
fflush(filep);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
outp = output_buffer;
}
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
void flush_output_stderr(void)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
{
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fputs(output_buffer, outf);
fflush(outf);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
outp = output_buffer;
}
void format_all_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools turbostat: add summary option turbostat -s cuts down on the amount of output, per user request. also treak some output whitespace and the man page. Signed-off-by: Len Brown <len.brown@intel.com>
2012-02-07 07:37:16 +08:00
static int printed;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools turbostat: add summary option turbostat -s cuts down on the amount of output, per user request. also treak some output whitespace and the man page. Signed-off-by: Len Brown <len.brown@intel.com>
2012-02-07 07:37:16 +08:00
if (!printed || !summary_only)
print_header();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (topo.num_cpus > 1)
format_counters(&average.threads, &average.cores,
&average.packages);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools turbostat: add summary option turbostat -s cuts down on the amount of output, per user request. also treak some output whitespace and the man page. Signed-off-by: Len Brown <len.brown@intel.com>
2012-02-07 07:37:16 +08:00
printed = 1;
if (summary_only)
return;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
for_all_cpus(format_counters, t, c, p);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
#define DELTA_WRAP32(new, old) \
if (new > old) { \
old = new - old; \
} else { \
old = 0x100000000 + new - old; \
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
void
delta_package(struct pkg_data *new, struct pkg_data *old)
{
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
if (do_skl_residency) {
old->pkg_wtd_core_c0 = new->pkg_wtd_core_c0 - old->pkg_wtd_core_c0;
old->pkg_any_core_c0 = new->pkg_any_core_c0 - old->pkg_any_core_c0;
old->pkg_any_gfxe_c0 = new->pkg_any_gfxe_c0 - old->pkg_any_gfxe_c0;
old->pkg_both_core_gfxe_c0 = new->pkg_both_core_gfxe_c0 - old->pkg_both_core_gfxe_c0;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
old->pc2 = new->pc2 - old->pc2;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc3)
old->pc3 = new->pc3 - old->pc3;
if (do_pc6)
old->pc6 = new->pc6 - old->pc6;
if (do_pc7)
old->pc7 = new->pc7 - old->pc7;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
old->pc8 = new->pc8 - old->pc8;
old->pc9 = new->pc9 - old->pc9;
old->pc10 = new->pc10 - old->pc10;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
old->pkg_temp_c = new->pkg_temp_c;
tools/power turbostat: work around RC6 counter wrap Sometimes the rc6 sysfs counter spontaneously resets, causing turbostat prints a very large number as it tries to calcuate % = 100 * (old - new) / interval When we see (old > new), print ***.**% instead of a bogus huge number. Note that this detection is not fool-proof, as the counter could reset several times and still result in new > old. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:16:00 +08:00
/* flag an error when rc6 counter resets/wraps */
if (old->gfx_rc6_ms > new->gfx_rc6_ms)
old->gfx_rc6_ms = -1;
else
old->gfx_rc6_ms = new->gfx_rc6_ms - old->gfx_rc6_ms;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
old->gfx_mhz = new->gfx_mhz;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
DELTA_WRAP32(new->energy_pkg, old->energy_pkg);
DELTA_WRAP32(new->energy_cores, old->energy_cores);
DELTA_WRAP32(new->energy_gfx, old->energy_gfx);
DELTA_WRAP32(new->energy_dram, old->energy_dram);
DELTA_WRAP32(new->rapl_pkg_perf_status, old->rapl_pkg_perf_status);
DELTA_WRAP32(new->rapl_dram_perf_status, old->rapl_dram_perf_status);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
void
delta_core(struct core_data *new, struct core_data *old)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
old->c3 = new->c3 - old->c3;
old->c6 = new->c6 - old->c6;
old->c7 = new->c7 - old->c7;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
old->core_temp_c = new->core_temp_c;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat: fix large c1% issue Under some conditions, c1% was displayed as very large number, much higher than 100%. c1% is not measured, it is derived as "that, which is left over" from other counters. However, the other counters are not collected atomically, and so it is possible for c1% to be calaculagted as a small negative number -- displayed as very large positive. There was a check for mperf vs tsc for this already, but it needed to also include the other counters that are used to calculate c1. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-14 09:31:46 +08:00
/*
* old = new - old
*/
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
void
delta_thread(struct thread_data *new, struct thread_data *old,
struct core_data *core_delta)
{
old->tsc = new->tsc - old->tsc;
/* check for TSC < 1 Mcycles over interval */
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (old->tsc < (1000 * 1000))
errx(-3, "Insanely slow TSC rate, TSC stops in idle?\n"
"You can disable all c-states by booting with \"idle=poll\"\n"
"or just the deep ones with \"processor.max_cstate=1\"");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
old->c1 = new->c1 - old->c1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: relax dependency on APERF_MSR While turbostat is significantly less useful on systems with no APERF_MSR, it seems more friendly to run on such systems and report what we can, rather than refusing to run. Update man page to reflect recent changes. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 14:33:58 +08:00
if (has_aperf) {
if ((new->aperf > old->aperf) && (new->mperf > old->mperf)) {
old->aperf = new->aperf - old->aperf;
old->mperf = new->mperf - old->mperf;
} else {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: relax dependency on APERF_MSR While turbostat is significantly less useful on systems with no APERF_MSR, it seems more friendly to run on such systems and report what we can, rather than refusing to run. Update man page to reflect recent changes. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 14:33:58 +08:00
if (!aperf_mperf_unstable) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%s: APERF or MPERF went backwards *\n", progname);
fprintf(outf, "* Frequency results do not cover entire interval *\n");
fprintf(outf, "* fix this by running Linux-2.6.30 or later *\n");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: relax dependency on APERF_MSR While turbostat is significantly less useful on systems with no APERF_MSR, it seems more friendly to run on such systems and report what we can, rather than refusing to run. Update man page to reflect recent changes. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 14:33:58 +08:00
aperf_mperf_unstable = 1;
}
/*
* mperf delta is likely a huge "positive" number
* can not use it for calculating c0 time
*/
skip_c0 = 1;
skip_c1 = 1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
if (use_c1_residency_msr) {
/*
* Some models have a dedicated C1 residency MSR,
* which should be more accurate than the derivation below.
*/
} else {
/*
* As counter collection is not atomic,
* it is possible for mperf's non-halted cycles + idle states
* to exceed TSC's all cycles: show c1 = 0% in that case.
*/
if ((old->mperf + core_delta->c3 + core_delta->c6 + core_delta->c7) > old->tsc)
old->c1 = 0;
else {
/* normal case, derive c1 */
old->c1 = old->tsc - old->mperf - core_delta->c3
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
- core_delta->c6 - core_delta->c7;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools/power: turbostat: fix large c1% issue Under some conditions, c1% was displayed as very large number, much higher than 100%. c1% is not measured, it is derived as "that, which is left over" from other counters. However, the other counters are not collected atomically, and so it is possible for c1% to be calaculagted as a small negative number -- displayed as very large positive. There was a check for mperf vs tsc for this already, but it needed to also include the other counters that are used to calculate c1. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-14 09:31:46 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (old->mperf == 0) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
if (debug > 1)
fprintf(outf, "cpu%d MPERF 0!\n", old->cpu_id);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
old->mperf = 1; /* divide by 0 protection */
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
old->extra_delta32 = new->extra_delta32 - old->extra_delta32;
old->extra_delta32 &= 0xFFFFFFFF;
old->extra_delta64 = new->extra_delta64 - old->extra_delta64;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/*
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
* Extra MSR is just a snapshot, simply copy latest w/o subtracting
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
*/
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
old->extra_msr32 = new->extra_msr32;
old->extra_msr64 = new->extra_msr64;
tools/power turbostat: display SMI count by default The SMI counter is popular -- so display it by default rather than requiring an option. What the heck, we've blown the 80 column budget on many systems already... Note that the value displayed is the delta during the measurement interval. The absolute value of the counter can still be seen with the generic 32-bit MSR option, ie. -m 0x34 Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-11 06:19:24 +08:00
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
if (do_irq)
old->irq_count = new->irq_count - old->irq_count;
tools/power turbostat: display SMI count by default The SMI counter is popular -- so display it by default rather than requiring an option. What the heck, we've blown the 80 column budget on many systems already... Note that the value displayed is the delta during the measurement interval. The absolute value of the counter can still be seen with the generic 32-bit MSR option, ie. -m 0x34 Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-11 06:19:24 +08:00
if (do_smi)
old->smi_count = new->smi_count - old->smi_count;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
int delta_cpu(struct thread_data *t, struct core_data *c,
struct pkg_data *p, struct thread_data *t2,
struct core_data *c2, struct pkg_data *p2)
{
/* calculate core delta only for 1st thread in core */
if (t->flags & CPU_IS_FIRST_THREAD_IN_CORE)
delta_core(c, c2);
/* always calculate thread delta */
delta_thread(t, t2, c2); /* c2 is core delta */
/* calculate package delta only for 1st core in package */
if (t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)
delta_package(p, p2);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
return 0;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
void clear_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
t->tsc = 0;
t->aperf = 0;
t->mperf = 0;
t->c1 = 0;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
t->extra_delta32 = 0;
t->extra_delta64 = 0;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
t->irq_count = 0;
t->smi_count = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* tells format_counters to dump all fields from this set */
t->flags = CPU_IS_FIRST_THREAD_IN_CORE | CPU_IS_FIRST_CORE_IN_PACKAGE;
c->c3 = 0;
c->c6 = 0;
c->c7 = 0;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
c->core_temp_c = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
p->pkg_wtd_core_c0 = 0;
p->pkg_any_core_c0 = 0;
p->pkg_any_gfxe_c0 = 0;
p->pkg_both_core_gfxe_c0 = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
p->pc2 = 0;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc3)
p->pc3 = 0;
if (do_pc6)
p->pc6 = 0;
if (do_pc7)
p->pc7 = 0;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
p->pc8 = 0;
p->pc9 = 0;
p->pc10 = 0;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
p->energy_pkg = 0;
p->energy_dram = 0;
p->energy_cores = 0;
p->energy_gfx = 0;
p->rapl_pkg_perf_status = 0;
p->rapl_dram_perf_status = 0;
p->pkg_temp_c = 0;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
p->gfx_rc6_ms = 0;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
p->gfx_mhz = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
int sum_counters(struct thread_data *t, struct core_data *c,
struct pkg_data *p)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
average.threads.tsc += t->tsc;
average.threads.aperf += t->aperf;
average.threads.mperf += t->mperf;
average.threads.c1 += t->c1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
average.threads.extra_delta32 += t->extra_delta32;
average.threads.extra_delta64 += t->extra_delta64;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
average.threads.irq_count += t->irq_count;
average.threads.smi_count += t->smi_count;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* sum per-core values only for 1st thread in core */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE))
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
average.cores.c3 += c->c3;
average.cores.c6 += c->c6;
average.cores.c7 += c->c7;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
average.cores.core_temp_c = MAX(average.cores.core_temp_c, c->core_temp_c);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* sum per-pkg values only for 1st core in pkg */
if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
return 0;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
if (do_skl_residency) {
average.packages.pkg_wtd_core_c0 += p->pkg_wtd_core_c0;
average.packages.pkg_any_core_c0 += p->pkg_any_core_c0;
average.packages.pkg_any_gfxe_c0 += p->pkg_any_gfxe_c0;
average.packages.pkg_both_core_gfxe_c0 += p->pkg_both_core_gfxe_c0;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
average.packages.pc2 += p->pc2;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc3)
average.packages.pc3 += p->pc3;
if (do_pc6)
average.packages.pc6 += p->pc6;
if (do_pc7)
average.packages.pc7 += p->pc7;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
average.packages.pc8 += p->pc8;
average.packages.pc9 += p->pc9;
average.packages.pc10 += p->pc10;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
average.packages.energy_pkg += p->energy_pkg;
average.packages.energy_dram += p->energy_dram;
average.packages.energy_cores += p->energy_cores;
average.packages.energy_gfx += p->energy_gfx;
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
average.packages.gfx_rc6_ms = p->gfx_rc6_ms;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
average.packages.gfx_mhz = p->gfx_mhz;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
average.packages.pkg_temp_c = MAX(average.packages.pkg_temp_c, p->pkg_temp_c);
average.packages.rapl_pkg_perf_status += p->rapl_pkg_perf_status;
average.packages.rapl_dram_perf_status += p->rapl_dram_perf_status;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return 0;
}
/*
* sum the counters for all cpus in the system
* compute the weighted average
*/
void compute_average(struct thread_data *t, struct core_data *c,
struct pkg_data *p)
{
clear_counters(&average.threads, &average.cores, &average.packages);
for_all_cpus(sum_counters, t, c, p);
average.threads.tsc /= topo.num_cpus;
average.threads.aperf /= topo.num_cpus;
average.threads.mperf /= topo.num_cpus;
average.threads.c1 /= topo.num_cpus;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
average.threads.extra_delta32 /= topo.num_cpus;
average.threads.extra_delta32 &= 0xFFFFFFFF;
average.threads.extra_delta64 /= topo.num_cpus;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
average.cores.c3 /= topo.num_cores;
average.cores.c6 /= topo.num_cores;
average.cores.c7 /= topo.num_cores;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
if (do_skl_residency) {
average.packages.pkg_wtd_core_c0 /= topo.num_packages;
average.packages.pkg_any_core_c0 /= topo.num_packages;
average.packages.pkg_any_gfxe_c0 /= topo.num_packages;
average.packages.pkg_both_core_gfxe_c0 /= topo.num_packages;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
average.packages.pc2 /= topo.num_packages;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc3)
average.packages.pc3 /= topo.num_packages;
if (do_pc6)
average.packages.pc6 /= topo.num_packages;
if (do_pc7)
average.packages.pc7 /= topo.num_packages;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
average.packages.pc8 /= topo.num_packages;
average.packages.pc9 /= topo.num_packages;
average.packages.pc10 /= topo.num_packages;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
static unsigned long long rdtsc(void)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
unsigned int low, high;
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
asm volatile("rdtsc" : "=a" (low), "=d" (high));
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return low | ((unsigned long long)high) << 32;
}
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/*
* get_counters(...)
* migrate to cpu
* acquire and record local counters for that cpu
*/
int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
int cpu = t->cpu_id;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
unsigned long long msr;
tools/power turbostat: detect and work around syscall jitter The accuracy of Bzy_Mhz and Busy% depend on reading the TSC, APERF, and MPERF close together in time. When there is a very short measurement interval, or a large system is profoundly idle, the changes in APERF and MPERF may be very small. They can be small enough that an expensive interrupt between reading APERF and MPERF can cause the APERF/MPERF ratio to become inaccurate, resulting in invalid calculation and display of Bzy_MHz. A dummy APERF read of APERF makes this problem much more rare. Apparently this 1st systemn call after exiting a long stretch of idle is when we typically see expensive timer interrupts that cause large jitter. For the cases that dummy APERF read fails to prevent, we compare the latency of the APERF and MPERF reads. If they differ by more than 2x, we re-issue them. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 16:11:29 +08:00
int aperf_mperf_retry_count = 0;
tools turbostat: reduce measurement overhead due to IPIs turbostat uses /dev/cpu/*/msr interface to read MSRs. For modern systems, it reads 10 MSR/CPU. This can be observed as 10 "Function Call Interrupts" per CPU per sample added to /proc/interrupts. This overhead is measurable on large idle systems, and as Yoquan Song pointed out, it can even trick cpuidle into thinking the system is busy. Here turbostat re-schedules itself in-turn to each CPU so that its MSR reads will always be local. This replaces the 10 "Function Call Interrupts" with a single "Rescheduling interrupt" per sample per CPU. On an idle 32-CPU system, this shifts some residency from the shallow c1 state to the deeper c7 state: # ./turbostat.old -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.29 2.29 0.95 0.02 0.00 98.77 20.23 0.00 77.41 0.00 0.25 1.24 2.29 0.98 0.02 0.00 98.75 20.34 0.03 77.74 0.00 0.27 1.22 2.29 0.54 0.00 0.00 99.18 20.64 0.00 77.70 0.00 0.26 1.22 2.29 1.22 0.00 0.00 98.52 20.22 0.00 77.74 0.00 0.26 1.38 2.29 0.78 0.02 0.00 98.95 20.51 0.05 77.56 0.00 ^C i# ./turbostat.new -s %c0 GHz TSC %c1 %c3 %c6 %c7 %pc2 %pc3 %pc6 %pc7 0.27 1.20 2.29 0.24 0.01 0.00 99.49 20.58 0.00 78.20 0.00 0.27 1.22 2.29 0.25 0.00 0.00 99.48 20.79 0.00 77.85 0.00 0.27 1.20 2.29 0.25 0.02 0.00 99.46 20.71 0.03 77.89 0.00 0.28 1.26 2.29 0.25 0.01 0.00 99.46 20.89 0.02 77.67 0.00 0.27 1.20 2.29 0.24 0.01 0.00 99.48 20.65 0.00 78.04 0.00 cc: Youquan Song <youquan.song@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 09:44:40 +08:00
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-09 11:38:05 +08:00
if (cpu_migrate(cpu)) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "Could not migrate to CPU %d\n", cpu);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return -1;
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-09 11:38:05 +08:00
}
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
tools/power turbostat: detect and work around syscall jitter The accuracy of Bzy_Mhz and Busy% depend on reading the TSC, APERF, and MPERF close together in time. When there is a very short measurement interval, or a large system is profoundly idle, the changes in APERF and MPERF may be very small. They can be small enough that an expensive interrupt between reading APERF and MPERF can cause the APERF/MPERF ratio to become inaccurate, resulting in invalid calculation and display of Bzy_MHz. A dummy APERF read of APERF makes this problem much more rare. Apparently this 1st systemn call after exiting a long stretch of idle is when we typically see expensive timer interrupts that cause large jitter. For the cases that dummy APERF read fails to prevent, we compare the latency of the APERF and MPERF reads. If they differ by more than 2x, we re-issue them. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 16:11:29 +08:00
retry:
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
t->tsc = rdtsc(); /* we are running on local CPU of interest */
if (has_aperf) {
tools/power turbostat: detect and work around syscall jitter The accuracy of Bzy_Mhz and Busy% depend on reading the TSC, APERF, and MPERF close together in time. When there is a very short measurement interval, or a large system is profoundly idle, the changes in APERF and MPERF may be very small. They can be small enough that an expensive interrupt between reading APERF and MPERF can cause the APERF/MPERF ratio to become inaccurate, resulting in invalid calculation and display of Bzy_MHz. A dummy APERF read of APERF makes this problem much more rare. Apparently this 1st systemn call after exiting a long stretch of idle is when we typically see expensive timer interrupts that cause large jitter. For the cases that dummy APERF read fails to prevent, we compare the latency of the APERF and MPERF reads. If they differ by more than 2x, we re-issue them. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 16:11:29 +08:00
unsigned long long tsc_before, tsc_between, tsc_after, aperf_time, mperf_time;
/*
* The TSC, APERF and MPERF must be read together for
* APERF/MPERF and MPERF/TSC to give accurate results.
*
* Unfortunately, APERF and MPERF are read by
* individual system call, so delays may occur
* between them. If the time to read them
* varies by a large amount, we re-read them.
*/
/*
* This initial dummy APERF read has been seen to
* reduce jitter in the subsequent reads.
*/
if (get_msr(cpu, MSR_IA32_APERF, &t->aperf))
return -3;
t->tsc = rdtsc(); /* re-read close to APERF */
tsc_before = t->tsc;
tools/power/x86/turbostat: share kernel MSR #defines Now that turbostat is built in the kernel tree, it can share MSR #defines with the kernel. Signed-off-by: Len Brown <len.brown@intel.com> Cc: x86@kernel.org
2012-10-31 13:29:52 +08:00
if (get_msr(cpu, MSR_IA32_APERF, &t->aperf))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return -3;
tools/power turbostat: detect and work around syscall jitter The accuracy of Bzy_Mhz and Busy% depend on reading the TSC, APERF, and MPERF close together in time. When there is a very short measurement interval, or a large system is profoundly idle, the changes in APERF and MPERF may be very small. They can be small enough that an expensive interrupt between reading APERF and MPERF can cause the APERF/MPERF ratio to become inaccurate, resulting in invalid calculation and display of Bzy_MHz. A dummy APERF read of APERF makes this problem much more rare. Apparently this 1st systemn call after exiting a long stretch of idle is when we typically see expensive timer interrupts that cause large jitter. For the cases that dummy APERF read fails to prevent, we compare the latency of the APERF and MPERF reads. If they differ by more than 2x, we re-issue them. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 16:11:29 +08:00
tsc_between = rdtsc();
tools/power/x86/turbostat: share kernel MSR #defines Now that turbostat is built in the kernel tree, it can share MSR #defines with the kernel. Signed-off-by: Len Brown <len.brown@intel.com> Cc: x86@kernel.org
2012-10-31 13:29:52 +08:00
if (get_msr(cpu, MSR_IA32_MPERF, &t->mperf))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return -4;
tools/power turbostat: detect and work around syscall jitter The accuracy of Bzy_Mhz and Busy% depend on reading the TSC, APERF, and MPERF close together in time. When there is a very short measurement interval, or a large system is profoundly idle, the changes in APERF and MPERF may be very small. They can be small enough that an expensive interrupt between reading APERF and MPERF can cause the APERF/MPERF ratio to become inaccurate, resulting in invalid calculation and display of Bzy_MHz. A dummy APERF read of APERF makes this problem much more rare. Apparently this 1st systemn call after exiting a long stretch of idle is when we typically see expensive timer interrupts that cause large jitter. For the cases that dummy APERF read fails to prevent, we compare the latency of the APERF and MPERF reads. If they differ by more than 2x, we re-issue them. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 16:11:29 +08:00
tsc_after = rdtsc();
aperf_time = tsc_between - tsc_before;
mperf_time = tsc_after - tsc_between;
/*
* If the system call latency to read APERF and MPERF
* differ by more than 2x, then try again.
*/
if ((aperf_time > (2 * mperf_time)) || (mperf_time > (2 * aperf_time))) {
aperf_mperf_retry_count++;
if (aperf_mperf_retry_count < 5)
goto retry;
else
warnx("cpu%d jitter %lld %lld",
cpu, aperf_time, mperf_time);
}
aperf_mperf_retry_count = 0;
tools/power turbostat: KNL workaround for %Busy and Avg_MHz KNL increments APERF and MPERF every 1024 clocks. This is compliant with the architecture specification, which requires that only the ratio of APERF/MPERF need be valid. However, turbostat takes advantage of the fact that these two MSRs increment every un-halted clock at the actual and base frequency: AVG_MHz = APERF_delta/measurement_interval %Busy = MPERF_delta/TSC_delta This quirk is needed for these calculations to also work on KNL, which would otherwise show a value 1024x smaller than expected. Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-14 19:31:00 +08:00
t->aperf = t->aperf * aperf_mperf_multiplier;
t->mperf = t->mperf * aperf_mperf_multiplier;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
if (do_irq)
t->irq_count = irqs_per_cpu[cpu];
tools/power turbostat: display SMI count by default The SMI counter is popular -- so display it by default rather than requiring an option. What the heck, we've blown the 80 column budget on many systems already... Note that the value displayed is the delta during the measurement interval. The absolute value of the counter can still be seen with the generic 32-bit MSR option, ie. -m 0x34 Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-11 06:19:24 +08:00
if (do_smi) {
if (get_msr(cpu, MSR_SMI_COUNT, &msr))
return -5;
t->smi_count = msr & 0xFFFFFFFF;
}
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
if (extra_delta_offset32) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, extra_delta_offset32, &msr))
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
return -5;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
t->extra_delta32 = msr & 0xFFFFFFFF;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
}
if (extra_delta_offset64)
if (get_msr(cpu, extra_delta_offset64, &t->extra_delta64))
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
return -5;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
if (extra_msr_offset32) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, extra_msr_offset32, &msr))
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
return -5;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
t->extra_msr32 = msr & 0xFFFFFFFF;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
}
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
if (extra_msr_offset64)
if (get_msr(cpu, extra_msr_offset64, &t->extra_msr64))
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return -5;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
if (use_c1_residency_msr) {
if (get_msr(cpu, MSR_CORE_C1_RES, &t->c1))
return -6;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* collect core counters only for 1st thread in core */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE))
return 0;
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
if (do_nhm_cstates && !do_slm_cstates && !do_knl_cstates) {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (get_msr(cpu, MSR_CORE_C3_RESIDENCY, &c->c3))
return -6;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
}
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
if (do_nhm_cstates && !do_knl_cstates) {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (get_msr(cpu, MSR_CORE_C6_RESIDENCY, &c->c6))
return -7;
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
} else if (do_knl_cstates) {
if (get_msr(cpu, MSR_KNL_CORE_C6_RESIDENCY, &c->c6))
return -7;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
if (do_snb_cstates)
if (get_msr(cpu, MSR_CORE_C7_RESIDENCY, &c->c7))
return -8;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (do_dts) {
if (get_msr(cpu, MSR_IA32_THERM_STATUS, &msr))
return -9;
c->core_temp_c = tcc_activation_temp - ((msr >> 16) & 0x7F);
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* collect package counters only for 1st core in package */
if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
return 0;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
if (do_skl_residency) {
if (get_msr(cpu, MSR_PKG_WEIGHTED_CORE_C0_RES, &p->pkg_wtd_core_c0))
return -10;
if (get_msr(cpu, MSR_PKG_ANY_CORE_C0_RES, &p->pkg_any_core_c0))
return -11;
if (get_msr(cpu, MSR_PKG_ANY_GFXE_C0_RES, &p->pkg_any_gfxe_c0))
return -12;
if (get_msr(cpu, MSR_PKG_BOTH_CORE_GFXE_C0_RES, &p->pkg_both_core_gfxe_c0))
return -13;
}
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc3)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (get_msr(cpu, MSR_PKG_C3_RESIDENCY, &p->pc3))
return -9;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc6)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (get_msr(cpu, MSR_PKG_C6_RESIDENCY, &p->pc6))
return -10;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc2)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (get_msr(cpu, MSR_PKG_C2_RESIDENCY, &p->pc2))
return -11;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
if (do_pc7)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (get_msr(cpu, MSR_PKG_C7_RESIDENCY, &p->pc7))
return -12;
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
if (do_c8_c9_c10) {
if (get_msr(cpu, MSR_PKG_C8_RESIDENCY, &p->pc8))
return -13;
if (get_msr(cpu, MSR_PKG_C9_RESIDENCY, &p->pc9))
return -13;
if (get_msr(cpu, MSR_PKG_C10_RESIDENCY, &p->pc10))
return -13;
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (do_rapl & RAPL_PKG) {
if (get_msr(cpu, MSR_PKG_ENERGY_STATUS, &msr))
return -13;
p->energy_pkg = msr & 0xFFFFFFFF;
}
if (do_rapl & RAPL_CORES) {
if (get_msr(cpu, MSR_PP0_ENERGY_STATUS, &msr))
return -14;
p->energy_cores = msr & 0xFFFFFFFF;
}
if (do_rapl & RAPL_DRAM) {
if (get_msr(cpu, MSR_DRAM_ENERGY_STATUS, &msr))
return -15;
p->energy_dram = msr & 0xFFFFFFFF;
}
if (do_rapl & RAPL_GFX) {
if (get_msr(cpu, MSR_PP1_ENERGY_STATUS, &msr))
return -16;
p->energy_gfx = msr & 0xFFFFFFFF;
}
if (do_rapl & RAPL_PKG_PERF_STATUS) {
if (get_msr(cpu, MSR_PKG_PERF_STATUS, &msr))
return -16;
p->rapl_pkg_perf_status = msr & 0xFFFFFFFF;
}
if (do_rapl & RAPL_DRAM_PERF_STATUS) {
if (get_msr(cpu, MSR_DRAM_PERF_STATUS, &msr))
return -16;
p->rapl_dram_perf_status = msr & 0xFFFFFFFF;
}
if (do_ptm) {
if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_STATUS, &msr))
return -17;
p->pkg_temp_c = tcc_activation_temp - ((msr >> 16) & 0x7F);
}
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
if (do_gfx_rc6_ms)
p->gfx_rc6_ms = gfx_cur_rc6_ms;
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
if (do_gfx_mhz)
p->gfx_mhz = gfx_cur_mhz;
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
/*
* MSR_PKG_CST_CONFIG_CONTROL decoding for pkg_cstate_limit:
* If you change the values, note they are used both in comparisons
* (>= PCL__7) and to index pkg_cstate_limit_strings[].
*/
#define PCLUKN 0 /* Unknown */
#define PCLRSV 1 /* Reserved */
#define PCL__0 2 /* PC0 */
#define PCL__1 3 /* PC1 */
#define PCL__2 4 /* PC2 */
#define PCL__3 5 /* PC3 */
#define PCL__4 6 /* PC4 */
#define PCL__6 7 /* PC6 */
#define PCL_6N 8 /* PC6 No Retention */
#define PCL_6R 9 /* PC6 Retention */
#define PCL__7 10 /* PC7 */
#define PCL_7S 11 /* PC7 Shrink */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
#define PCL__8 12 /* PC8 */
#define PCL__9 13 /* PC9 */
#define PCLUNL 14 /* Unlimited */
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
int pkg_cstate_limit = PCLUKN;
char *pkg_cstate_limit_strings[] = { "reserved", "unknown", "pc0", "pc1", "pc2",
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
"pc3", "pc4", "pc6", "pc6n", "pc6r", "pc7", "pc7s", "pc8", "pc9", "unlimited"};
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
tools/power turbostat: correct dumped pkg-cstate-limit value HSW expanded MSR_PKG_CST_CONFIG_CONTROL.Package-C-State-Limit, from bits[2:0] used by previous implementations, to [3:0]. The value 1000b is unlimited, and is used by BDW and SKL too. Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
int nhm_pkg_cstate_limits[16] = {PCL__0, PCL__1, PCL__3, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV};
int snb_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCL_6N, PCL_6R, PCL__7, PCL_7S, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV};
int hsw_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCL__3, PCL__6, PCL__7, PCL_7S, PCL__8, PCL__9, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV};
int slv_pkg_cstate_limits[16] = {PCL__0, PCL__1, PCLRSV, PCLRSV, PCL__4, PCLRSV, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV};
int amt_pkg_cstate_limits[16] = {PCL__0, PCL__1, PCL__2, PCLRSV, PCLRSV, PCLRSV, PCL__6, PCL__7, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV};
int phi_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCL_6N, PCL_6R, PCLRSV, PCLRSV, PCLRSV, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV};
tools/power turbostat: initial BXT support Broxton has a lot in common with SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:56 +08:00
int bxt_pkg_cstate_limits[16] = {PCL__0, PCL__2, PCLUNL, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV, PCLRSV};
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
tools/power turbostat: SKL: Adjust for TSC difference from base frequency On a Skylake with 1500MHz base frequency, the TSC runs at 1512MHz. This is because the TSC is no longer in the n*100 MHz BCLK domain, but is now in the m*24MHz crystal clock domain. (24 MHz * 63 = 1512 MHz) This adds error to several calculations in turbostat, unless the TSC sample sizes are adjusted for this difference. Note that calculations in the time domain are immune from this issue, as the timing sub-system has already calibrated the TSC against a known wall clock. AVG_MHz = APERF_delta/measurement_interval need no adjustment. APERF_delta is in the BCLK domain, and measurement_interval is in the time domain. TSC_MHz = TSC_delta/measurement_interval needs no adjustment -- as we really do want to report the actual measured TSC delta here, and measurement_interval is in the accurate time domain. %Busy = MPERF_delta/TSC_delta needs adjustment to use TSC_BCLK_DOMAIN_delta. TSC_BCLK_DOMAIN_delta = TSC_delta * base_hz / tsc_hz Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval need adjustment as above. No other metrics in turbostat need to be adjusted. Before: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 24.84 2216 1512 0 2191 98.73 2219 1514 2 0 0.01 2130 1512 1 9 0.43 2016 1512 3 2 0.08 2016 1512 After: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 25.05 2198 1512 0 2190 99.62 2199 1512 2 0 0.01 2152 1512 1 9 0.46 2000 1512 3 2 0.10 2000 1512 Note that in this example, the "Before" Bzy_MHz was reported as exceeding the 2200 max turbo rate. Also, even a pinned spin loop would not be reported as over 99% busy. Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-26 12:12:38 +08:00
static void
calculate_tsc_tweak()
{
tsc_tweak = base_hz / tsc_hz;
}
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
static void
dump_nhm_platform_info(void)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
unsigned long long msr;
unsigned int ratio;
tools/power turbostat: use new name for MSR_PLATFORM_INFO MSR_PLATFORM_INFO is the new name for MSR_NHM_PLATFORM_INFO no functional change Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-11-12 15:42:31 +08:00
get_msr(base_cpu, MSR_PLATFORM_INFO, &msr);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_PLATFORM_INFO: 0x%08llx\n", base_cpu, msr);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
ratio = (msr >> 40) & 0xFF;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max efficiency frequency\n",
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 8) & 0xFF;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz base frequency\n",
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
ratio, bclk, ratio * bclk);
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
get_msr(base_cpu, MSR_IA32_POWER_CTL, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_IA32_POWER_CTL: 0x%08llx (C1E auto-promotion: %sabled)\n",
tools/power turbostat: cpu0 is no longer hard-coded, so update output The --debug option reads a number of per-package MSRs. Previously we explicitly read them on cpu0, but recently turbostat changed to read them on the current "base_cpu". Update the print-out to reflect base_cpu, rather than the hard-coded cpu0. Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 00:27:21 +08:00
base_cpu, msr, msr & 0x2 ? "EN" : "DIS");
tools/power turbostat: decode MSR_IA32_POWER_CTL When verbose is enabled, print the C1E-Enable bit in MSR_IA32_POWER_CTL. also delete some redundant tests on the verbose variable. Signed-off-by: Len Brown <len.brown@intel.com>
2013-02-01 04:22:15 +08:00
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
return;
}
static void
dump_hsw_turbo_ratio_limits(void)
{
unsigned long long msr;
unsigned int ratio;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT2, &msr);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT2: 0x%08llx\n", base_cpu, msr);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
ratio = (msr >> 8) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 18 active cores\n",
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 0) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 17 active cores\n",
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
ratio, bclk, ratio * bclk);
return;
}
static void
dump_ivt_turbo_ratio_limits(void)
{
unsigned long long msr;
unsigned int ratio;
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT1, &msr);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT1: 0x%08llx\n", base_cpu, msr);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio = (msr >> 56) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 16 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 48) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 15 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 40) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 14 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 32) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 13 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 24) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 12 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 16) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 11 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 8) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 10 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 0) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 9 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
return;
}
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
static void
dump_nhm_turbo_ratio_limits(void)
{
unsigned long long msr;
unsigned int ratio;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
get_msr(base_cpu, MSR_TURBO_RATIO_LIMIT, &msr);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT: 0x%08llx\n", base_cpu, msr);
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio = (msr >> 56) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 8 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 48) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 7 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 40) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 6 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 32) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 5 active cores\n",
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
ratio, bclk, ratio * bclk);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
ratio = (msr >> 24) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 4 active cores\n",
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 16) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 3 active cores\n",
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 8) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 2 active cores\n",
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
ratio, bclk, ratio * bclk);
ratio = (msr >> 0) & 0xFF;
if (ratio)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d * %.0f = %.0f MHz max turbo 1 active cores\n",
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
ratio, bclk, ratio * bclk);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
return;
}
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
static void
dump_knl_turbo_ratio_limits(void)
{
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 21:55:22 +08:00
const unsigned int buckets_no = 7;
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
unsigned long long msr;
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 21:55:22 +08:00
int delta_cores, delta_ratio;
int i, b_nr;
unsigned int cores[buckets_no];
unsigned int ratio[buckets_no];
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
get_msr(base_cpu, MSR_NHM_TURBO_RATIO_LIMIT, &msr);
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_TURBO_RATIO_LIMIT: 0x%08llx\n",
tools/power turbostat: cpu0 is no longer hard-coded, so update output The --debug option reads a number of per-package MSRs. Previously we explicitly read them on cpu0, but recently turbostat changed to read them on the current "base_cpu". Update the print-out to reflect base_cpu, rather than the hard-coded cpu0. Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 00:27:21 +08:00
base_cpu, msr);
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
/**
* Turbo encoding in KNL is as follows:
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 21:55:22 +08:00
* [0] -- Reserved
* [7:1] -- Base value of number of active cores of bucket 1.
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
* [15:8] -- Base value of freq ratio of bucket 1.
* [20:16] -- +ve delta of number of active cores of bucket 2.
* i.e. active cores of bucket 2 =
* active cores of bucket 1 + delta
* [23:21] -- Negative delta of freq ratio of bucket 2.
* i.e. freq ratio of bucket 2 =
* freq ratio of bucket 1 - delta
* [28:24]-- +ve delta of number of active cores of bucket 3.
* [31:29]-- -ve delta of freq ratio of bucket 3.
* [36:32]-- +ve delta of number of active cores of bucket 4.
* [39:37]-- -ve delta of freq ratio of bucket 4.
* [44:40]-- +ve delta of number of active cores of bucket 5.
* [47:45]-- -ve delta of freq ratio of bucket 5.
* [52:48]-- +ve delta of number of active cores of bucket 6.
* [55:53]-- -ve delta of freq ratio of bucket 6.
* [60:56]-- +ve delta of number of active cores of bucket 7.
* [63:61]-- -ve delta of freq ratio of bucket 7.
*/
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 21:55:22 +08:00
b_nr = 0;
cores[b_nr] = (msr & 0xFF) >> 1;
ratio[b_nr] = (msr >> 8) & 0xFF;
for (i = 16; i < 64; i += 8) {
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
delta_cores = (msr >> i) & 0x1F;
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 21:55:22 +08:00
delta_ratio = (msr >> (i + 5)) & 0x7;
cores[b_nr + 1] = cores[b_nr] + delta_cores;
ratio[b_nr + 1] = ratio[b_nr] - delta_ratio;
b_nr++;
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
}
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 21:55:22 +08:00
for (i = buckets_no - 1; i >= 0; i--)
if (i > 0 ? ratio[i] != ratio[i - 1] : 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf,
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
"%d * %.0f = %.0f MHz max turbo %d active cores\n",
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 21:55:22 +08:00
ratio[i], bclk, ratio[i] * bclk, cores[i]);
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
}
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
static void
dump_nhm_cst_cfg(void)
{
unsigned long long msr;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
get_msr(base_cpu, MSR_NHM_SNB_PKG_CST_CFG_CTL, &msr);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
#define SNB_C1_AUTO_UNDEMOTE (1UL << 27)
#define SNB_C3_AUTO_UNDEMOTE (1UL << 28)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_NHM_SNB_PKG_CST_CFG_CTL: 0x%08llx", base_cpu, msr);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, " (%s%s%s%s%slocked: pkg-cstate-limit=%d: %s)\n",
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
(msr & SNB_C3_AUTO_UNDEMOTE) ? "UNdemote-C3, " : "",
(msr & SNB_C1_AUTO_UNDEMOTE) ? "UNdemote-C1, " : "",
(msr & NHM_C3_AUTO_DEMOTE) ? "demote-C3, " : "",
(msr & NHM_C1_AUTO_DEMOTE) ? "demote-C1, " : "",
(msr & (1 << 15)) ? "" : "UN",
tools/power turbostat: correct output for MSR_NHM_SNB_PKG_CST_CFG_CTL dump MSR_NHM_SNB_PKG_CST_CFG_CTL: 0x1e008008 (...pkg-cstate-limit=0: unlimited) should print as MSR_NHM_SNB_PKG_CST_CFG_CTL: 0x1e008008 (...pkg-cstate-limit=8: unlimited) Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 15:21:22 +08:00
(unsigned int)msr & 0xF,
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
pkg_cstate_limit_strings[pkg_cstate_limit]);
return;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
static void
dump_config_tdp(void)
{
unsigned long long msr;
get_msr(base_cpu, MSR_CONFIG_TDP_NOMINAL, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_CONFIG_TDP_NOMINAL: 0x%08llx", base_cpu, msr);
tools/power turbostat: bugfix: TDP MSRs print bits fixing MSR_CONFIG_TDP_NOMINAL: should print all 8 bits of base_ratio (bit 0:7) 0xFF MSR_CONFIG_TDP_LEVEL_1: should print all 15 bits of PKG_MIN_PWR_LVL1 (bit 48:62) 0x7FFF should print all 15 bits of PKG_MAX_PWR_LVL1 (bit 32:46) 0x7FFF should print all 8 bits of LVL1_RATIO (bit 16:23) 0xFF should print all 15 bits of PKG_TDP_LVL1 (bit 0:14) 0x7FFF And the same modification to MSR_CONFIG_TDP_LEVEL_2. MSR_TURBO_ACTIVATION_RATIO: should print all 8 bits of MAX_NON_TURBO_RATIO (bit 0:7) 0xFF Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-13 21:09:31 +08:00
fprintf(outf, " (base_ratio=%d)\n", (unsigned int)msr & 0xFF);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
get_msr(base_cpu, MSR_CONFIG_TDP_LEVEL_1, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_CONFIG_TDP_LEVEL_1: 0x%08llx (", base_cpu, msr);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
if (msr) {
tools/power turbostat: bugfix: TDP MSRs print bits fixing MSR_CONFIG_TDP_NOMINAL: should print all 8 bits of base_ratio (bit 0:7) 0xFF MSR_CONFIG_TDP_LEVEL_1: should print all 15 bits of PKG_MIN_PWR_LVL1 (bit 48:62) 0x7FFF should print all 15 bits of PKG_MAX_PWR_LVL1 (bit 32:46) 0x7FFF should print all 8 bits of LVL1_RATIO (bit 16:23) 0xFF should print all 15 bits of PKG_TDP_LVL1 (bit 0:14) 0x7FFF And the same modification to MSR_CONFIG_TDP_LEVEL_2. MSR_TURBO_ACTIVATION_RATIO: should print all 8 bits of MAX_NON_TURBO_RATIO (bit 0:7) 0xFF Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-13 21:09:31 +08:00
fprintf(outf, "PKG_MIN_PWR_LVL1=%d ", (unsigned int)(msr >> 48) & 0x7FFF);
fprintf(outf, "PKG_MAX_PWR_LVL1=%d ", (unsigned int)(msr >> 32) & 0x7FFF);
fprintf(outf, "LVL1_RATIO=%d ", (unsigned int)(msr >> 16) & 0xFF);
fprintf(outf, "PKG_TDP_LVL1=%d", (unsigned int)(msr) & 0x7FFF);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
}
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, ")\n");
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
get_msr(base_cpu, MSR_CONFIG_TDP_LEVEL_2, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_CONFIG_TDP_LEVEL_2: 0x%08llx (", base_cpu, msr);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
if (msr) {
tools/power turbostat: bugfix: TDP MSRs print bits fixing MSR_CONFIG_TDP_NOMINAL: should print all 8 bits of base_ratio (bit 0:7) 0xFF MSR_CONFIG_TDP_LEVEL_1: should print all 15 bits of PKG_MIN_PWR_LVL1 (bit 48:62) 0x7FFF should print all 15 bits of PKG_MAX_PWR_LVL1 (bit 32:46) 0x7FFF should print all 8 bits of LVL1_RATIO (bit 16:23) 0xFF should print all 15 bits of PKG_TDP_LVL1 (bit 0:14) 0x7FFF And the same modification to MSR_CONFIG_TDP_LEVEL_2. MSR_TURBO_ACTIVATION_RATIO: should print all 8 bits of MAX_NON_TURBO_RATIO (bit 0:7) 0xFF Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-13 21:09:31 +08:00
fprintf(outf, "PKG_MIN_PWR_LVL2=%d ", (unsigned int)(msr >> 48) & 0x7FFF);
fprintf(outf, "PKG_MAX_PWR_LVL2=%d ", (unsigned int)(msr >> 32) & 0x7FFF);
fprintf(outf, "LVL2_RATIO=%d ", (unsigned int)(msr >> 16) & 0xFF);
fprintf(outf, "PKG_TDP_LVL2=%d", (unsigned int)(msr) & 0x7FFF);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
}
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, ")\n");
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
get_msr(base_cpu, MSR_CONFIG_TDP_CONTROL, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_CONFIG_TDP_CONTROL: 0x%08llx (", base_cpu, msr);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
if ((msr) & 0x3)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "TDP_LEVEL=%d ", (unsigned int)(msr) & 0x3);
fprintf(outf, " lock=%d", (unsigned int)(msr >> 31) & 1);
fprintf(outf, ")\n");
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
get_msr(base_cpu, MSR_TURBO_ACTIVATION_RATIO, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_TURBO_ACTIVATION_RATIO: 0x%08llx (", base_cpu, msr);
tools/power turbostat: bugfix: TDP MSRs print bits fixing MSR_CONFIG_TDP_NOMINAL: should print all 8 bits of base_ratio (bit 0:7) 0xFF MSR_CONFIG_TDP_LEVEL_1: should print all 15 bits of PKG_MIN_PWR_LVL1 (bit 48:62) 0x7FFF should print all 15 bits of PKG_MAX_PWR_LVL1 (bit 32:46) 0x7FFF should print all 8 bits of LVL1_RATIO (bit 16:23) 0xFF should print all 15 bits of PKG_TDP_LVL1 (bit 0:14) 0x7FFF And the same modification to MSR_CONFIG_TDP_LEVEL_2. MSR_TURBO_ACTIVATION_RATIO: should print all 8 bits of MAX_NON_TURBO_RATIO (bit 0:7) 0xFF Signed-off-by: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-13 21:09:31 +08:00
fprintf(outf, "MAX_NON_TURBO_RATIO=%d", (unsigned int)(msr) & 0xFF);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, " lock=%d", (unsigned int)(msr >> 31) & 1);
fprintf(outf, ")\n");
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
}
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:55 +08:00
unsigned int irtl_time_units[] = {1, 32, 1024, 32768, 1048576, 33554432, 0, 0 };
void print_irtl(void)
{
unsigned long long msr;
get_msr(base_cpu, MSR_PKGC3_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC3_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
get_msr(base_cpu, MSR_PKGC6_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC6_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
get_msr(base_cpu, MSR_PKGC7_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC7_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
if (!do_irtl_hsw)
return;
get_msr(base_cpu, MSR_PKGC8_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC8_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
get_msr(base_cpu, MSR_PKGC9_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC9_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
get_msr(base_cpu, MSR_PKGC10_IRTL, &msr);
fprintf(outf, "cpu%d: MSR_PKGC10_IRTL: 0x%08llx (", base_cpu, msr);
fprintf(outf, "%svalid, %lld ns)\n", msr & (1 << 15) ? "" : "NOT",
(msr & 0x3FF) * irtl_time_units[(msr >> 10) & 0x3]);
}
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
void free_fd_percpu(void)
{
int i;
for (i = 0; i < topo.max_cpu_num; ++i) {
if (fd_percpu[i] != 0)
close(fd_percpu[i]);
}
free(fd_percpu);
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
void free_all_buffers(void)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
CPU_FREE(cpu_present_set);
cpu_present_set = NULL;
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
cpu_present_setsize = 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
CPU_FREE(cpu_affinity_set);
cpu_affinity_set = NULL;
cpu_affinity_setsize = 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
free(thread_even);
free(core_even);
free(package_even);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
thread_even = NULL;
core_even = NULL;
package_even = NULL;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
free(thread_odd);
free(core_odd);
free(package_odd);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
thread_odd = NULL;
core_odd = NULL;
package_odd = NULL;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
free(output_buffer);
output_buffer = NULL;
outp = NULL;
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
free_fd_percpu();
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
free(irq_column_2_cpu);
free(irqs_per_cpu);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
turbostat: Factor out common function to open file and exit on failure Several different functions in turbostat contain the same pattern of opening a file and exiting on failure. Factor out a common fopen_or_die function for that. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:17 +08:00
/*
* Open a file, and exit on failure
*/
FILE *fopen_or_die(const char *path, const char *mode)
{
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
FILE *filep = fopen(path, mode);
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (!filep)
err(1, "%s: open failed", path);
turbostat: Factor out common function to open file and exit on failure Several different functions in turbostat contain the same pattern of opening a file and exiting on failure. Factor out a common fopen_or_die function for that. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:17 +08:00
return filep;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/*
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
* Parse a file containing a single int.
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
*/
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
int parse_int_file(const char *fmt, ...)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
va_list args;
char path[PATH_MAX];
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
FILE *filep;
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
int value;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
va_start(args, fmt);
vsnprintf(path, sizeof(path), fmt, args);
va_end(args);
turbostat: Factor out common function to open file and exit on failure Several different functions in turbostat contain the same pattern of opening a file and exiting on failure. Factor out a common fopen_or_die function for that. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:17 +08:00
filep = fopen_or_die(path, "r");
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (fscanf(filep, "%d", &value) != 1)
err(1, "%s: failed to parse number from file", path);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
fclose(filep);
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
return value;
}
/*
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
* get_cpu_position_in_core(cpu)
* return the position of the CPU among its HT siblings in the core
* return -1 if the sibling is not in list
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
*/
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
int get_cpu_position_in_core(int cpu)
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
{
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
char path[64];
FILE *filep;
int this_cpu;
char character;
int i;
sprintf(path,
"/sys/devices/system/cpu/cpu%d/topology/thread_siblings_list",
cpu);
filep = fopen(path, "r");
if (filep == NULL) {
perror(path);
exit(1);
}
for (i = 0; i < topo.num_threads_per_core; i++) {
fscanf(filep, "%d", &this_cpu);
if (this_cpu == cpu) {
fclose(filep);
return i;
}
/* Account for no separator after last thread*/
if (i != (topo.num_threads_per_core - 1))
fscanf(filep, "%c", &character);
}
fclose(filep);
return -1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/*
* cpu_is_first_core_in_package(cpu)
* return 1 if given CPU is 1st core in package
*/
int cpu_is_first_core_in_package(int cpu)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
return cpu == parse_int_file("/sys/devices/system/cpu/cpu%d/topology/core_siblings_list", cpu);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
int get_physical_package_id(int cpu)
{
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/physical_package_id", cpu);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
int get_core_id(int cpu)
{
turbostat: Add a helper to parse a single int out of a file Many different chunks of code in turbostat open a file, parse a single int out of it, and close it. Factor that out into a common function. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:16 +08:00
return parse_int_file("/sys/devices/system/cpu/cpu%d/topology/core_id", cpu);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int get_num_ht_siblings(int cpu)
{
char path[80];
FILE *filep;
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
int sib1;
int matches = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
char character;
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
char str[100];
char *ch;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
sprintf(path, "/sys/devices/system/cpu/cpu%d/topology/thread_siblings_list", cpu);
turbostat: Factor out common function to open file and exit on failure Several different functions in turbostat contain the same pattern of opening a file and exiting on failure. Factor out a common fopen_or_die function for that. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:17 +08:00
filep = fopen_or_die(path, "r");
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/*
* file format:
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
* A ',' separated or '-' separated set of numbers
* (eg 1-2 or 1,3,4,5)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
*/
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
fscanf(filep, "%d%c\n", &sib1, &character);
fseek(filep, 0, SEEK_SET);
fgets(str, 100, filep);
ch = strchr(str, character);
while (ch != NULL) {
matches++;
ch = strchr(ch+1, character);
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
fclose(filep);
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
return matches+1;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
/*
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
* run func(thread, core, package) in topology order
* skip non-present cpus
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
*/
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int for_all_cpus_2(int (func)(struct thread_data *, struct core_data *,
struct pkg_data *, struct thread_data *, struct core_data *,
struct pkg_data *), struct thread_data *thread_base,
struct core_data *core_base, struct pkg_data *pkg_base,
struct thread_data *thread_base2, struct core_data *core_base2,
struct pkg_data *pkg_base2)
{
int retval, pkg_no, core_no, thread_no;
for (pkg_no = 0; pkg_no < topo.num_packages; ++pkg_no) {
for (core_no = 0; core_no < topo.num_cores_per_pkg; ++core_no) {
for (thread_no = 0; thread_no <
topo.num_threads_per_core; ++thread_no) {
struct thread_data *t, *t2;
struct core_data *c, *c2;
struct pkg_data *p, *p2;
t = GET_THREAD(thread_base, thread_no, core_no, pkg_no);
if (cpu_is_not_present(t->cpu_id))
continue;
t2 = GET_THREAD(thread_base2, thread_no, core_no, pkg_no);
c = GET_CORE(core_base, core_no, pkg_no);
c2 = GET_CORE(core_base2, core_no, pkg_no);
p = GET_PKG(pkg_base, pkg_no);
p2 = GET_PKG(pkg_base2, pkg_no);
retval = func(t, c, p, t2, c2, p2);
if (retval)
return retval;
}
}
}
return 0;
}
/*
* run func(cpu) on every cpu in /proc/stat
* return max_cpu number
*/
int for_all_proc_cpus(int (func)(int))
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
FILE *fp;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int cpu_num;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
int retval;
turbostat: Factor out common function to open file and exit on failure Several different functions in turbostat contain the same pattern of opening a file and exiting on failure. Factor out a common fopen_or_die function for that. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:17 +08:00
fp = fopen_or_die(proc_stat, "r");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
retval = fscanf(fp, "cpu %*d %*d %*d %*d %*d %*d %*d %*d %*d %*d\n");
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (retval != 0)
err(1, "%s: failed to parse format", proc_stat);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
while (1) {
retval = fscanf(fp, "cpu%u %*d %*d %*d %*d %*d %*d %*d %*d %*d %*d\n", &cpu_num);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (retval != 1)
break;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
retval = func(cpu_num);
if (retval) {
fclose(fp);
return(retval);
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
fclose(fp);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
void re_initialize(void)
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
free_all_buffers();
setup_all_buffers();
printf("turbostat: re-initialized with num_cpus %d\n", topo.num_cpus);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
/*
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
* count_cpus()
* remember the last one seen, it will be the max
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
*/
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int count_cpus(int cpu)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (topo.max_cpu_num < cpu)
topo.max_cpu_num = cpu;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
topo.num_cpus += 1;
return 0;
}
int mark_cpu_present(int cpu)
{
CPU_SET_S(cpu, cpu_present_setsize, cpu_present_set);
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
return 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
/*
* snapshot_proc_interrupts()
*
* read and record summary of /proc/interrupts
*
* return 1 if config change requires a restart, else return 0
*/
int snapshot_proc_interrupts(void)
{
static FILE *fp;
int column, retval;
if (fp == NULL)
fp = fopen_or_die("/proc/interrupts", "r");
else
rewind(fp);
/* read 1st line of /proc/interrupts to get cpu* name for each column */
for (column = 0; column < topo.num_cpus; ++column) {
int cpu_number;
retval = fscanf(fp, " CPU%d", &cpu_number);
if (retval != 1)
break;
if (cpu_number > topo.max_cpu_num) {
warn("/proc/interrupts: cpu%d: > %d", cpu_number, topo.max_cpu_num);
return 1;
}
irq_column_2_cpu[column] = cpu_number;
irqs_per_cpu[cpu_number] = 0;
}
/* read /proc/interrupt count lines and sum up irqs per cpu */
while (1) {
int column;
char buf[64];
retval = fscanf(fp, " %s:", buf); /* flush irq# "N:" */
if (retval != 1)
break;
/* read the count per cpu */
for (column = 0; column < topo.num_cpus; ++column) {
int cpu_number, irq_count;
retval = fscanf(fp, " %d", &irq_count);
if (retval != 1)
break;
cpu_number = irq_column_2_cpu[column];
irqs_per_cpu[cpu_number] += irq_count;
}
while (getc(fp) != '\n')
; /* flush interrupt description */
}
return 0;
}
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
/*
* snapshot_gfx_rc6_ms()
*
* record snapshot of
* /sys/class/drm/card0/power/rc6_residency_ms
*
* return 1 if config change requires a restart, else return 0
*/
int snapshot_gfx_rc6_ms(void)
{
FILE *fp;
int retval;
fp = fopen_or_die("/sys/class/drm/card0/power/rc6_residency_ms", "r");
retval = fscanf(fp, "%lld", &gfx_cur_rc6_ms);
if (retval != 1)
err(1, "GFX rc6");
fclose(fp);
return 0;
}
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
/*
* snapshot_gfx_mhz()
*
* record snapshot of
* /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz
*
* return 1 if config change requires a restart, else return 0
*/
int snapshot_gfx_mhz(void)
{
static FILE *fp;
int retval;
if (fp == NULL)
fp = fopen_or_die("/sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz", "r");
else
rewind(fp);
retval = fscanf(fp, "%d", &gfx_cur_mhz);
if (retval != 1)
err(1, "GFX MHz");
return 0;
}
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
/*
* snapshot /proc and /sys files
*
* return 1 if configuration restart needed, else return 0
*/
int snapshot_proc_sysfs_files(void)
{
if (snapshot_proc_interrupts())
return 1;
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
if (do_gfx_rc6_ms)
snapshot_gfx_rc6_ms();
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
if (do_gfx_mhz)
snapshot_gfx_mhz();
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
return 0;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
void turbostat_loop()
{
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
int retval;
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-09 11:38:05 +08:00
int restarted = 0;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
restart:
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-09 11:38:05 +08:00
restarted++;
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
snapshot_proc_sysfs_files();
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
retval = for_all_cpus(get_counters, EVEN_COUNTERS);
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 12:08:19 +08:00
if (retval < -1) {
exit(retval);
} else if (retval == -1) {
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-09 11:38:05 +08:00
if (restarted > 1) {
exit(retval);
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
re_initialize();
goto restart;
}
tools/power turbostat: prevent infinite loop on migration error path Turbostat assumed if it can't migrate to a CPU, then the CPU must have gone off-line and turbostat should re-initialize with the new topology. But if turbostat can not migrate because it is restricted by a cpuset, then it will fail to migrate even after re-initialization, resulting in an infinite loop. Spit out a warning when we can't migrate and endure only 2 re-initialize cycles in a row before giving up and exiting. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-09 11:38:05 +08:00
restarted = 0;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
gettimeofday(&tv_even, (struct timezone *)NULL);
while (1) {
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
if (for_all_proc_cpus(cpu_is_not_present)) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
re_initialize();
goto restart;
}
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 11:44:48 +08:00
nanosleep(&interval_ts, NULL);
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
if (snapshot_proc_sysfs_files())
goto restart;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
retval = for_all_cpus(get_counters, ODD_COUNTERS);
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 12:08:19 +08:00
if (retval < -1) {
exit(retval);
} else if (retval == -1) {
tools turbostat: harden against cpu online/offline Sometimes users have turbostat running in interval mode when they take processors offline/online. Previously, turbostat would survive, but not gracefully. Tighten up the error checking so turbostat notices changesn sooner, and print just 1 line on change: turbostat: re-initialized with num_cpus %d Signed-off-by: Len Brown <len.brown@intel.com>
2012-03-30 10:19:58 +08:00
re_initialize();
goto restart;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
gettimeofday(&tv_odd, (struct timezone *)NULL);
timersub(&tv_odd, &tv_even, &tv_delta);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
for_all_cpus_2(delta_cpu, ODD_COUNTERS, EVEN_COUNTERS);
compute_average(EVEN_COUNTERS);
format_all_counters(EVEN_COUNTERS);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
flush_output_stdout();
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 11:44:48 +08:00
nanosleep(&interval_ts, NULL);
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
if (snapshot_proc_sysfs_files())
goto restart;
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
retval = for_all_cpus(get_counters, EVEN_COUNTERS);
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 12:08:19 +08:00
if (retval < -1) {
exit(retval);
} else if (retval == -1) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
re_initialize();
goto restart;
}
gettimeofday(&tv_even, (struct timezone *)NULL);
timersub(&tv_even, &tv_odd, &tv_delta);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
for_all_cpus_2(delta_cpu, EVEN_COUNTERS, ODD_COUNTERS);
compute_average(ODD_COUNTERS);
format_all_counters(ODD_COUNTERS);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
flush_output_stdout();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
}
void check_dev_msr()
{
struct stat sb;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
char pathname[32];
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
sprintf(pathname, "/dev/cpu/%d/msr", base_cpu);
if (stat(pathname, &sb))
tools/power turbostat: modprobe msr, if needed Some distros (Ubuntu) ship the msr driver as a module. If turbosat is run as root on those systems, and discovers that there is no /dev/cpu/cpu0/msr, it will now "modprobe msr" for the user. If not root, the modprobe attempt will fail, and turbostat will exit as before: turbostat: no /dev/cpu/0/msr, Try "# modprobe msr" : No such file or directory Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-25 04:37:35 +08:00
if (system("/sbin/modprobe msr > /dev/null 2>&1"))
err(-5, "no /dev/cpu/0/msr, Try \"# modprobe msr\" ");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
void check_permissions()
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
struct __user_cap_header_struct cap_header_data;
cap_user_header_t cap_header = &cap_header_data;
struct __user_cap_data_struct cap_data_data;
cap_user_data_t cap_data = &cap_data_data;
extern int capget(cap_user_header_t hdrp, cap_user_data_t datap);
int do_exit = 0;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
char pathname[32];
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
/* check for CAP_SYS_RAWIO */
cap_header->pid = getpid();
cap_header->version = _LINUX_CAPABILITY_VERSION;
if (capget(cap_header, cap_data) < 0)
err(-6, "capget(2) failed");
if ((cap_data->effective & (1 << CAP_SYS_RAWIO)) == 0) {
do_exit++;
warnx("capget(CAP_SYS_RAWIO) failed,"
" try \"# setcap cap_sys_rawio=ep %s\"", progname);
}
/* test file permissions */
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
sprintf(pathname, "/dev/cpu/%d/msr", base_cpu);
if (euidaccess(pathname, R_OK)) {
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
do_exit++;
warn("/dev/cpu/0/msr open failed, try chown or chmod +r /dev/cpu/*/msr");
}
/* if all else fails, thell them to be root */
if (do_exit)
if (getuid() != 0)
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
warnx("... or simply run as root");
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
if (do_exit)
exit(-6);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
/*
* NHM adds support for additional MSRs:
*
* MSR_SMI_COUNT 0x00000034
*
tools/power turbostat: use new name for MSR_PLATFORM_INFO MSR_PLATFORM_INFO is the new name for MSR_NHM_PLATFORM_INFO no functional change Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-11-12 15:42:31 +08:00
* MSR_PLATFORM_INFO 0x000000ce
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
* MSR_NHM_SNB_PKG_CST_CFG_CTL 0x000000e2
*
* MSR_PKG_C3_RESIDENCY 0x000003f8
* MSR_PKG_C6_RESIDENCY 0x000003f9
* MSR_CORE_C3_RESIDENCY 0x000003fc
* MSR_CORE_C6_RESIDENCY 0x000003fd
*
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
* Side effect:
* sets global pkg_cstate_limit to decode MSR_NHM_SNB_PKG_CST_CFG_CTL
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
*/
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
int probe_nhm_msrs(unsigned int family, unsigned int model)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
unsigned long long msr;
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-20 10:37:40 +08:00
unsigned int base_ratio;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
int *pkg_cstate_limits;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (!genuine_intel)
return 0;
if (family != 6)
return 0;
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-20 10:37:40 +08:00
bclk = discover_bclk(family, model);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
switch (model) {
case 0x1A: /* Core i7, Xeon 5500 series - Bloomfield, Gainstown NHM-EP */
case 0x1E: /* Core i7 and i5 Processor - Clarksfield, Lynnfield, Jasper Forest */
case 0x1F: /* Core i7 and i5 Processor - Nehalem */
case 0x25: /* Westmere Client - Clarkdale, Arrandale */
case 0x2C: /* Westmere EP - Gulftown */
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
case 0x2E: /* Nehalem-EX Xeon - Beckton */
case 0x2F: /* Westmere-EX Xeon - Eagleton */
pkg_cstate_limits = nhm_pkg_cstate_limits;
break;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
case 0x2A: /* SNB */
case 0x2D: /* SNB Xeon */
tools turbostat: recognize and run properly on IVB Signed-off-by: Len Brown <len.brown@intel.com>
2011-11-18 16:32:01 +08:00
case 0x3A: /* IVB */
tools/power turbostat: run on IVB Xeon This fix is required to run on IVB Xeon, which previously had an incorrect cpuid model number listed. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-27 06:11:31 +08:00
case 0x3E: /* IVB Xeon */
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
pkg_cstate_limits = snb_pkg_cstate_limits;
break;
tools/power turbostat: support Haswell This patch enables turbostat to run properly on the next-generation Intel(R) Microarchitecture, code named "Haswell" (HSW). HSW supports the BCLK and counters found in SNB. Signed-off-by: Len Brown <len.brown@intel.com>
2013-01-08 14:26:07 +08:00
case 0x3C: /* HSW */
turbostat: run on HSX Haswell Xeon has slightly different RAPL support than client HSW, which prevented the previous version of turbostat from running on HSX. Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-03 15:19:19 +08:00
case 0x3F: /* HSX */
tools/power turbostat: support Haswell This patch enables turbostat to run properly on the next-generation Intel(R) Microarchitecture, code named "Haswell" (HSW). HSW supports the BCLK and counters found in SNB. Signed-off-by: Len Brown <len.brown@intel.com>
2013-01-08 14:26:07 +08:00
case 0x45: /* HSW */
tools/power turbostat: additional Haswell CPU-id There is an additional HSW CPU-id, 0x46, which has C-states exactly like CPU-id 0x45. Signed-off-by: Len Brown <len.brown@intel.com>
2013-03-15 22:58:02 +08:00
case 0x46: /* HSW */
tools/power turbostat: Run on Broadwell Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-28 12:28:53 +08:00
case 0x3D: /* BDW */
tools/power turbostat: support additional Broadwell model Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-11 04:38:04 +08:00
case 0x47: /* BDW */
tools/power turbostat: Run on Broadwell Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-28 12:28:53 +08:00
case 0x4F: /* BDX */
case 0x56: /* BDX-DE */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
case 0x4E: /* SKL */
case 0x5E: /* SKL */
tools/power turbostat: initial KBL support KBL is similar to SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:59 +08:00
case 0x8E: /* KBL */
case 0x9E: /* KBL */
tools/power turbostat: initial SKX support SKX has a lot in common with HSX Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:58 +08:00
case 0x55: /* SKX */
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
pkg_cstate_limits = hsw_pkg_cstate_limits;
break;
case 0x37: /* BYT */
case 0x4D: /* AVN */
pkg_cstate_limits = slv_pkg_cstate_limits;
break;
case 0x4C: /* AMT */
pkg_cstate_limits = amt_pkg_cstate_limits;
break;
case 0x57: /* PHI */
pkg_cstate_limits = phi_pkg_cstate_limits;
break;
tools/power turbostat: initial BXT support Broxton has a lot in common with SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:56 +08:00
case 0x5C: /* BXT */
pkg_cstate_limits = bxt_pkg_cstate_limits;
break;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
default:
return 0;
}
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
get_msr(base_cpu, MSR_NHM_SNB_PKG_CST_CFG_CTL, &msr);
tools/power turbostat: correct dumped pkg-cstate-limit value HSW expanded MSR_PKG_CST_CONFIG_CONTROL.Package-C-State-Limit, from bits[2:0] used by previous implementations, to [3:0]. The value 1000b is unlimited, and is used by BDW and SKL too. Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
pkg_cstate_limit = pkg_cstate_limits[msr & 0xF];
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
tools/power turbostat: use new name for MSR_PLATFORM_INFO MSR_PLATFORM_INFO is the new name for MSR_NHM_PLATFORM_INFO no functional change Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-11-12 15:42:31 +08:00
get_msr(base_cpu, MSR_PLATFORM_INFO, &msr);
tools/power turbostat: simplify Bzy_MHz calculation Bzy_MHz = TSC_delta*tsc_tweak/APERF_delta/MPERF_delta/measurement_interval becomes Bzy_MHz = base_mhz/APERF_delta/MPERF_delta on systems which support MSR_NHM_PLATFORM_INFO. base_mhz is calculated directly from the base_ratio reported in MSR_NHM_PLATFORM_INFO * bclk, and bclk is discovered via MSR or cpuid. This reduces the dependency of Bzy_MHz calculation on the TSC. Previously, there were 4 TSC readings required in each caculation, the raw TSC delta combined with the measurement_interval. This also removes the "tsc_tweak" correction factor used when TSC runs on a different base clock from the CPU's bclk. After this change, tsc_tweak is used only for %Busy. The end-result should be a Bzy_MHz result slightly less prone to jitter. Signed-off-by: Len Brown <len.brown@intel.com>
2015-10-20 10:37:40 +08:00
base_ratio = (msr >> 8) & 0xFF;
base_hz = base_ratio * bclk * 1000000;
has_base_hz = 1;
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
return 1;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
int has_nhm_turbo_ratio_limit(unsigned int family, unsigned int model)
{
switch (model) {
/* Nehalem compatible, but do not include turbo-ratio limit support */
case 0x2E: /* Nehalem-EX Xeon - Beckton */
case 0x2F: /* Westmere-EX Xeon - Eagleton */
tools/power turbostat: Intel Xeon x200: fix turbo-ratio decoding Following changes have been made: - changed MSR_NHM_TURBO_RATIO_LIMIT to MSR_TURBO_RATIO_LIMIT in debug print for consistency with Developer Manual - updated definition of bitfields in MSR_TURBO_RATIO_LIMIT and appropriate parsing code - added x200 to list of architectures that do not support Nahlem compatible definition of MSR_TURBO_RATIO_LIMIT register (x200 has the register but bits definition is custom) - fixed typo in code that parses MSR_TURBO_RATIO_LIMIT (logical instead of bitwise operator) - changed MSR_TURBO_RATIO_LIMIT parsing algorithm so the print out had the same order as implementations for other platforms Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 21:55:22 +08:00
case 0x57: /* PHI - Knights Landing (different MSR definition) */
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
return 0;
default:
return 1;
}
}
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
int has_ivt_turbo_ratio_limit(unsigned int family, unsigned int model)
{
if (!genuine_intel)
return 0;
if (family != 6)
return 0;
switch (model) {
case 0x3E: /* IVB Xeon */
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
case 0x3F: /* HSW Xeon */
return 1;
default:
return 0;
}
}
int has_hsw_turbo_ratio_limit(unsigned int family, unsigned int model)
{
if (!genuine_intel)
return 0;
if (family != 6)
return 0;
switch (model) {
case 0x3F: /* HSW Xeon */
tools/power turbostat: print more turbo-limit information The "turbo-limit" is the maximum opportunistic processor speed, assuming no electrical or thermal constraints. For a given processor, the turbo-limit varies, depending on the number of active cores. Generally, there is more opportunity when fewer cores are active. Under the "-v" verbose option, turbostat would print the turbo-limits for the four cases of 1 to 4 cores active. Expand that capability to cover the cases of turbo opportunities with up to 16 cores active. Note that not all hardware platforms supply this information, and that sometimes a valid limit may be specified for a core which is not actually present. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-21 12:01:31 +08:00
return 1;
default:
return 0;
}
}
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
int has_knl_turbo_ratio_limit(unsigned int family, unsigned int model)
{
if (!genuine_intel)
return 0;
if (family != 6)
return 0;
switch (model) {
case 0x57: /* Knights Landing */
return 1;
default:
return 0;
}
}
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
int has_config_tdp(unsigned int family, unsigned int model)
{
if (!genuine_intel)
return 0;
if (family != 6)
return 0;
switch (model) {
case 0x3A: /* IVB */
case 0x3C: /* HSW */
case 0x3F: /* HSX */
case 0x45: /* HSW */
case 0x46: /* HSW */
case 0x3D: /* BDW */
case 0x47: /* BDW */
case 0x4F: /* BDX */
case 0x56: /* BDX-DE */
case 0x4E: /* SKL */
case 0x5E: /* SKL */
tools/power turbostat: initial KBL support KBL is similar to SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:59 +08:00
case 0x8E: /* KBL */
case 0x9E: /* KBL */
tools/power turbostat: initial SKX support SKX has a lot in common with HSX Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:58 +08:00
case 0x55: /* SKX */
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
case 0x57: /* Knights Landing */
return 1;
default:
return 0;
}
}
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
static void
tools/power turbostat: fix various build warnings When building with gcc 6 we're getting various build warnings that just require some trivial function declaration and call fixes: turbostat.c: In function ‘dump_cstate_pstate_config_info’: turbostat.c:1973:1: warning: type of ‘family’ defaults to ‘int’ dump_cstate_pstate_config_info(family, model) turbostat.c:1973:1: warning: type of ‘model’ defaults to ‘int’ turbostat.c: In function ‘get_tdp’: turbostat.c:2145:8: warning: type of ‘model’ defaults to ‘int’ double get_tdp(model) turbostat.c: In function ‘perf_limit_reasons_probe’: turbostat.c:2259:6: warning: type of ‘family’ defaults to ‘int’ void perf_limit_reasons_probe(family, model) turbostat.c:2259:6: warning: type of ‘model’ defaults to ‘int’ Signed-off-by: Colin Ian King <colin.king@canonical.com> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/n/tip-wbicer8n0s9qe6ql8h9x478e@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-03-02 21:50:25 +08:00
dump_cstate_pstate_config_info(unsigned int family, unsigned int model)
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
{
if (!do_nhm_platform_info)
return;
dump_nhm_platform_info();
if (has_hsw_turbo_ratio_limit(family, model))
dump_hsw_turbo_ratio_limits();
if (has_ivt_turbo_ratio_limit(family, model))
dump_ivt_turbo_ratio_limits();
if (has_nhm_turbo_ratio_limit(family, model))
dump_nhm_turbo_ratio_limits();
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
if (has_knl_turbo_ratio_limit(family, model))
dump_knl_turbo_ratio_limits();
tools/power turbostat: dump CONFIG_TDP Config TDP is a feature that allows parts to be configured for different thermal limits after they have left the factory. This can have an effect on the operation of the part, particularly in determiniing... Max Non-turbo Ratio Turbo Activation Ratio Signed-off-by: Len Brown <len.brown@intel.com>
2015-06-18 04:23:45 +08:00
if (has_config_tdp(family, model))
dump_config_tdp();
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
dump_nhm_cst_cfg();
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
/*
* print_epb()
* Decode the ENERGY_PERF_BIAS MSR
*/
int print_epb(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
char *epb_string;
int cpu;
if (!has_epb)
return 0;
cpu = t->cpu_id;
/* EPB is per-package */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
return 0;
if (cpu_migrate(cpu)) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "Could not migrate to CPU %d\n", cpu);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return -1;
}
if (get_msr(cpu, MSR_IA32_ENERGY_PERF_BIAS, &msr))
return 0;
tools/power turbostat: correctly decode of ENERGY_PERFORMANCE_BIAS When EPB is 0xF, turbosat was incorrectly describing it as "custom" instead of calling it "powersave": < cpu0: MSR_IA32_ENERGY_PERF_BIAS: 0x0000000f (custom) > cpu0: MSR_IA32_ENERGY_PERF_BIAS: 0x0000000f (powersave) Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-27 00:19:37 +08:00
switch (msr & 0xF) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
case ENERGY_PERF_BIAS_PERFORMANCE:
epb_string = "performance";
break;
case ENERGY_PERF_BIAS_NORMAL:
epb_string = "balanced";
break;
case ENERGY_PERF_BIAS_POWERSAVE:
epb_string = "powersave";
break;
default:
epb_string = "custom";
break;
}
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_IA32_ENERGY_PERF_BIAS: 0x%08llx (%s)\n", cpu, msr, epb_string);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return 0;
}
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
/*
* print_hwp()
* Decode the MSR_HWP_CAPABILITIES
*/
int print_hwp(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
int cpu;
if (!has_hwp)
return 0;
cpu = t->cpu_id;
/* MSR_HWP_CAPABILITIES is per-package */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
return 0;
if (cpu_migrate(cpu)) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "Could not migrate to CPU %d\n", cpu);
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
return -1;
}
if (get_msr(cpu, MSR_PM_ENABLE, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_PM_ENABLE: 0x%08llx (%sHWP)\n",
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
cpu, msr, (msr & (1 << 0)) ? "" : "No-");
/* MSR_PM_ENABLE[1] == 1 if HWP is enabled and MSRs visible */
if ((msr & (1 << 0)) == 0)
return 0;
if (get_msr(cpu, MSR_HWP_CAPABILITIES, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_HWP_CAPABILITIES: 0x%08llx "
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
"(high 0x%x guar 0x%x eff 0x%x low 0x%x)\n",
cpu, msr,
(unsigned int)HWP_HIGHEST_PERF(msr),
(unsigned int)HWP_GUARANTEED_PERF(msr),
(unsigned int)HWP_MOSTEFFICIENT_PERF(msr),
(unsigned int)HWP_LOWEST_PERF(msr));
if (get_msr(cpu, MSR_HWP_REQUEST, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_HWP_REQUEST: 0x%08llx "
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
"(min 0x%x max 0x%x des 0x%x epp 0x%x window 0x%x pkg 0x%x)\n",
cpu, msr,
(unsigned int)(((msr) >> 0) & 0xff),
(unsigned int)(((msr) >> 8) & 0xff),
(unsigned int)(((msr) >> 16) & 0xff),
(unsigned int)(((msr) >> 24) & 0xff),
(unsigned int)(((msr) >> 32) & 0xff3),
(unsigned int)(((msr) >> 42) & 0x1));
if (has_hwp_pkg) {
if (get_msr(cpu, MSR_HWP_REQUEST_PKG, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_HWP_REQUEST_PKG: 0x%08llx "
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
"(min 0x%x max 0x%x des 0x%x epp 0x%x window 0x%x)\n",
cpu, msr,
(unsigned int)(((msr) >> 0) & 0xff),
(unsigned int)(((msr) >> 8) & 0xff),
(unsigned int)(((msr) >> 16) & 0xff),
(unsigned int)(((msr) >> 24) & 0xff),
(unsigned int)(((msr) >> 32) & 0xff3));
}
if (has_hwp_notify) {
if (get_msr(cpu, MSR_HWP_INTERRUPT, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_HWP_INTERRUPT: 0x%08llx "
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
"(%s_Guaranteed_Perf_Change, %s_Excursion_Min)\n",
cpu, msr,
((msr) & 0x1) ? "EN" : "Dis",
((msr) & 0x2) ? "EN" : "Dis");
}
if (get_msr(cpu, MSR_HWP_STATUS, &msr))
return 0;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_HWP_STATUS: 0x%08llx "
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
"(%sGuaranteed_Perf_Change, %sExcursion_Min)\n",
cpu, msr,
((msr) & 0x1) ? "" : "No-",
((msr) & 0x2) ? "" : "No-");
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return 0;
}
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
/*
* print_perf_limit()
*/
int print_perf_limit(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
int cpu;
cpu = t->cpu_id;
/* per-package */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
return 0;
if (cpu_migrate(cpu)) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "Could not migrate to CPU %d\n", cpu);
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
return -1;
}
if (do_core_perf_limit_reasons) {
get_msr(cpu, MSR_CORE_PERF_LIMIT_REASONS, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_CORE_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr);
fprintf(outf, " (Active: %s%s%s%s%s%s%s%s%s%s%s%s%s%s)",
tools/power turbostat: update PERF_LIMIT_REASONS decoding cosmetic only. order the decoding of MSR_PERF_LIMIT_REASONS bits from MSB to LSB -- which you notice when more than 1 bit is set and you are, say, comparing the output to the documentation... Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-14 04:30:57 +08:00
(msr & 1 << 15) ? "bit15, " : "",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
(msr & 1 << 14) ? "bit14, " : "",
tools/power turbostat: update PERF_LIMIT_REASONS decoding cosmetic only. order the decoding of MSR_PERF_LIMIT_REASONS bits from MSB to LSB -- which you notice when more than 1 bit is set and you are, say, comparing the output to the documentation... Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-14 04:30:57 +08:00
(msr & 1 << 13) ? "Transitions, " : "",
(msr & 1 << 12) ? "MultiCoreTurbo, " : "",
(msr & 1 << 11) ? "PkgPwrL2, " : "",
(msr & 1 << 10) ? "PkgPwrL1, " : "",
(msr & 1 << 9) ? "CorePwr, " : "",
(msr & 1 << 8) ? "Amps, " : "",
(msr & 1 << 6) ? "VR-Therm, " : "",
(msr & 1 << 5) ? "Auto-HWP, " : "",
(msr & 1 << 4) ? "Graphics, " : "",
(msr & 1 << 2) ? "bit2, " : "",
(msr & 1 << 1) ? "ThermStatus, " : "",
(msr & 1 << 0) ? "PROCHOT, " : "");
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, " (Logged: %s%s%s%s%s%s%s%s%s%s%s%s%s%s)\n",
tools/power turbostat: update PERF_LIMIT_REASONS decoding cosmetic only. order the decoding of MSR_PERF_LIMIT_REASONS bits from MSB to LSB -- which you notice when more than 1 bit is set and you are, say, comparing the output to the documentation... Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-14 04:30:57 +08:00
(msr & 1 << 31) ? "bit31, " : "",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
(msr & 1 << 30) ? "bit30, " : "",
tools/power turbostat: update PERF_LIMIT_REASONS decoding cosmetic only. order the decoding of MSR_PERF_LIMIT_REASONS bits from MSB to LSB -- which you notice when more than 1 bit is set and you are, say, comparing the output to the documentation... Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-14 04:30:57 +08:00
(msr & 1 << 29) ? "Transitions, " : "",
(msr & 1 << 28) ? "MultiCoreTurbo, " : "",
(msr & 1 << 27) ? "PkgPwrL2, " : "",
(msr & 1 << 26) ? "PkgPwrL1, " : "",
(msr & 1 << 25) ? "CorePwr, " : "",
(msr & 1 << 24) ? "Amps, " : "",
(msr & 1 << 22) ? "VR-Therm, " : "",
(msr & 1 << 21) ? "Auto-HWP, " : "",
(msr & 1 << 20) ? "Graphics, " : "",
(msr & 1 << 18) ? "bit18, " : "",
(msr & 1 << 17) ? "ThermStatus, " : "",
(msr & 1 << 16) ? "PROCHOT, " : "");
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
}
if (do_gfx_perf_limit_reasons) {
get_msr(cpu, MSR_GFX_PERF_LIMIT_REASONS, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_GFX_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr);
fprintf(outf, " (Active: %s%s%s%s%s%s%s%s)",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
(msr & 1 << 0) ? "PROCHOT, " : "",
(msr & 1 << 1) ? "ThermStatus, " : "",
(msr & 1 << 4) ? "Graphics, " : "",
(msr & 1 << 6) ? "VR-Therm, " : "",
(msr & 1 << 8) ? "Amps, " : "",
(msr & 1 << 9) ? "GFXPwr, " : "",
(msr & 1 << 10) ? "PkgPwrL1, " : "",
(msr & 1 << 11) ? "PkgPwrL2, " : "");
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, " (Logged: %s%s%s%s%s%s%s%s)\n",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
(msr & 1 << 16) ? "PROCHOT, " : "",
(msr & 1 << 17) ? "ThermStatus, " : "",
(msr & 1 << 20) ? "Graphics, " : "",
(msr & 1 << 22) ? "VR-Therm, " : "",
(msr & 1 << 24) ? "Amps, " : "",
(msr & 1 << 25) ? "GFXPwr, " : "",
(msr & 1 << 26) ? "PkgPwrL1, " : "",
(msr & 1 << 27) ? "PkgPwrL2, " : "");
}
if (do_ring_perf_limit_reasons) {
get_msr(cpu, MSR_RING_PERF_LIMIT_REASONS, &msr);
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_RING_PERF_LIMIT_REASONS, 0x%08llx", cpu, msr);
fprintf(outf, " (Active: %s%s%s%s%s%s)",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
(msr & 1 << 0) ? "PROCHOT, " : "",
(msr & 1 << 1) ? "ThermStatus, " : "",
(msr & 1 << 6) ? "VR-Therm, " : "",
(msr & 1 << 8) ? "Amps, " : "",
(msr & 1 << 10) ? "PkgPwrL1, " : "",
(msr & 1 << 11) ? "PkgPwrL2, " : "");
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, " (Logged: %s%s%s%s%s%s)\n",
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
(msr & 1 << 16) ? "PROCHOT, " : "",
(msr & 1 << 17) ? "ThermStatus, " : "",
(msr & 1 << 22) ? "VR-Therm, " : "",
(msr & 1 << 24) ? "Amps, " : "",
(msr & 1 << 26) ? "PkgPwrL1, " : "",
(msr & 1 << 27) ? "PkgPwrL2, " : "");
}
return 0;
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
#define RAPL_POWER_GRANULARITY 0x7FFF /* 15 bit power granularity */
#define RAPL_TIME_GRANULARITY 0x3F /* 6 bit time granularity */
tools/power turbostat: fix various build warnings When building with gcc 6 we're getting various build warnings that just require some trivial function declaration and call fixes: turbostat.c: In function ‘dump_cstate_pstate_config_info’: turbostat.c:1973:1: warning: type of ‘family’ defaults to ‘int’ dump_cstate_pstate_config_info(family, model) turbostat.c:1973:1: warning: type of ‘model’ defaults to ‘int’ turbostat.c: In function ‘get_tdp’: turbostat.c:2145:8: warning: type of ‘model’ defaults to ‘int’ double get_tdp(model) turbostat.c: In function ‘perf_limit_reasons_probe’: turbostat.c:2259:6: warning: type of ‘family’ defaults to ‘int’ void perf_limit_reasons_probe(family, model) turbostat.c:2259:6: warning: type of ‘model’ defaults to ‘int’ Signed-off-by: Colin Ian King <colin.king@canonical.com> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/n/tip-wbicer8n0s9qe6ql8h9x478e@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-03-02 21:50:25 +08:00
double get_tdp(unsigned int model)
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
{
unsigned long long msr;
if (do_rapl & RAPL_PKG_POWER_INFO)
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
if (!get_msr(base_cpu, MSR_PKG_POWER_INFO, &msr))
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
return ((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units;
switch (model) {
case 0x37:
case 0x4D:
return 30.0;
default:
return 135.0;
}
}
tools/power turbostat: correct DRAM RAPL units on recent Xeon processors While not yet documented in the Software Developer's Manual, the data-sheet for modern Xeon states that DRAM RAPL ENERGY units are fixed at 15.3 uJ, rather than being discovered via MSR. Before this patch, DRAM energy on these products is over-stated by turbostat because the RAPL units are 4x larger. ref: "Xeon E5-2600 v3/E5-1600 v3 Datasheet Volume 2" http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e5-v3-datasheet-vol-2.pdf Signed-off-by: Andrey Semin <andrey.semin@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-12-05 13:07:00 +08:00
/*
* rapl_dram_energy_units_probe()
* Energy units are either hard-coded, or come from RAPL Energy Unit MSR.
*/
static double
rapl_dram_energy_units_probe(int model, double rapl_energy_units)
{
/* only called for genuine_intel, family 6 */
switch (model) {
case 0x3F: /* HSX */
case 0x4F: /* BDX */
case 0x56: /* BDX-DE */
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
case 0x57: /* KNL */
tools/power turbostat: correct DRAM RAPL units on recent Xeon processors While not yet documented in the Software Developer's Manual, the data-sheet for modern Xeon states that DRAM RAPL ENERGY units are fixed at 15.3 uJ, rather than being discovered via MSR. Before this patch, DRAM energy on these products is over-stated by turbostat because the RAPL units are 4x larger. ref: "Xeon E5-2600 v3/E5-1600 v3 Datasheet Volume 2" http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e5-v3-datasheet-vol-2.pdf Signed-off-by: Andrey Semin <andrey.semin@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-12-05 13:07:00 +08:00
return (rapl_dram_energy_units = 15.3 / 1000000);
default:
return (rapl_energy_units);
}
}
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
/*
* rapl_probe()
*
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
* sets do_rapl, rapl_power_units, rapl_energy_units, rapl_time_units
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
*/
void rapl_probe(unsigned int family, unsigned int model)
{
unsigned long long msr;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
unsigned int time_unit;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
double tdp;
if (!genuine_intel)
return;
if (family != 6)
return;
switch (model) {
case 0x2A:
case 0x3A:
tools/power turbostat: support Haswell This patch enables turbostat to run properly on the next-generation Intel(R) Microarchitecture, code named "Haswell" (HSW). HSW supports the BCLK and counters found in SNB. Signed-off-by: Len Brown <len.brown@intel.com>
2013-01-08 14:26:07 +08:00
case 0x3C: /* HSW */
case 0x45: /* HSW */
tools/power turbostat: additional Haswell CPU-id There is an additional HSW CPU-id, 0x46, which has C-states exactly like CPU-id 0x45. Signed-off-by: Len Brown <len.brown@intel.com>
2013-03-15 22:58:02 +08:00
case 0x46: /* HSW */
tools/power turbostat: Run on Broadwell Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-28 12:28:53 +08:00
case 0x3D: /* BDW */
tools/power turbostat: support additional Broadwell model Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-11 04:38:04 +08:00
case 0x47: /* BDW */
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
do_rapl = RAPL_PKG | RAPL_CORES | RAPL_CORE_POLICY | RAPL_GFX | RAPL_PKG_POWER_INFO;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
break;
tools/power turbostat: initial BXT support Broxton has a lot in common with SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:56 +08:00
case 0x5C: /* BXT */
do_rapl = RAPL_PKG | RAPL_PKG_POWER_INFO;
break;
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
case 0x4E: /* SKL */
case 0x5E: /* SKL */
tools/power turbostat: initial KBL support KBL is similar to SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:59 +08:00
case 0x8E: /* KBL */
case 0x9E: /* KBL */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
do_rapl = RAPL_PKG | RAPL_DRAM | RAPL_DRAM_PERF_STATUS | RAPL_PKG_PERF_STATUS | RAPL_PKG_POWER_INFO;
break;
turbostat: run on HSX Haswell Xeon has slightly different RAPL support than client HSW, which prevented the previous version of turbostat from running on HSX. Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-03 15:19:19 +08:00
case 0x3F: /* HSX */
tools/power turbostat: Run on Broadwell Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-28 12:28:53 +08:00
case 0x4F: /* BDX */
case 0x56: /* BDX-DE */
tools/power turbostat: initial SKX support SKX has a lot in common with HSX Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:58 +08:00
case 0x55: /* SKX */
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
case 0x57: /* KNL */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
do_rapl = RAPL_PKG | RAPL_DRAM | RAPL_DRAM_POWER_INFO | RAPL_DRAM_PERF_STATUS | RAPL_PKG_PERF_STATUS | RAPL_PKG_POWER_INFO;
turbostat: run on HSX Haswell Xeon has slightly different RAPL support than client HSW, which prevented the previous version of turbostat from running on HSX. Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-03 15:19:19 +08:00
break;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
case 0x2D:
case 0x3E:
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
do_rapl = RAPL_PKG | RAPL_CORES | RAPL_CORE_POLICY | RAPL_DRAM | RAPL_DRAM_POWER_INFO | RAPL_PKG_PERF_STATUS | RAPL_DRAM_PERF_STATUS | RAPL_PKG_POWER_INFO;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
break;
case 0x37: /* BYT */
case 0x4D: /* AVN */
do_rapl = RAPL_PKG | RAPL_CORES ;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
break;
default:
return;
}
/* units on package 0, verify later other packages match */
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
if (get_msr(base_cpu, MSR_RAPL_POWER_UNIT, &msr))
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return;
rapl_power_units = 1.0 / (1 << (msr & 0xF));
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
if (model == 0x37)
rapl_energy_units = 1.0 * (1 << (msr >> 8 & 0x1F)) / 1000000;
else
rapl_energy_units = 1.0 / (1 << (msr >> 8 & 0x1F));
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools/power turbostat: correct DRAM RAPL units on recent Xeon processors While not yet documented in the Software Developer's Manual, the data-sheet for modern Xeon states that DRAM RAPL ENERGY units are fixed at 15.3 uJ, rather than being discovered via MSR. Before this patch, DRAM energy on these products is over-stated by turbostat because the RAPL units are 4x larger. ref: "Xeon E5-2600 v3/E5-1600 v3 Datasheet Volume 2" http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e5-v3-datasheet-vol-2.pdf Signed-off-by: Andrey Semin <andrey.semin@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-12-05 13:07:00 +08:00
rapl_dram_energy_units = rapl_dram_energy_units_probe(model, rapl_energy_units);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
time_unit = msr >> 16 & 0xF;
if (time_unit == 0)
time_unit = 0xA;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
rapl_time_units = 1.0 / (1 << (time_unit));
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
tdp = get_tdp(model);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
rapl_joule_counter_range = 0xFFFFFFFF * rapl_energy_units / tdp;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "RAPL: %.0f sec. Joule Counter Range, at %.0f Watts\n", rapl_joule_counter_range, tdp);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return;
}
tools/power turbostat: fix various build warnings When building with gcc 6 we're getting various build warnings that just require some trivial function declaration and call fixes: turbostat.c: In function ‘dump_cstate_pstate_config_info’: turbostat.c:1973:1: warning: type of ‘family’ defaults to ‘int’ dump_cstate_pstate_config_info(family, model) turbostat.c:1973:1: warning: type of ‘model’ defaults to ‘int’ turbostat.c: In function ‘get_tdp’: turbostat.c:2145:8: warning: type of ‘model’ defaults to ‘int’ double get_tdp(model) turbostat.c: In function ‘perf_limit_reasons_probe’: turbostat.c:2259:6: warning: type of ‘family’ defaults to ‘int’ void perf_limit_reasons_probe(family, model) turbostat.c:2259:6: warning: type of ‘model’ defaults to ‘int’ Signed-off-by: Colin Ian King <colin.king@canonical.com> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/n/tip-wbicer8n0s9qe6ql8h9x478e@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-03-02 21:50:25 +08:00
void perf_limit_reasons_probe(unsigned int family, unsigned int model)
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
{
if (!genuine_intel)
return;
if (family != 6)
return;
switch (model) {
case 0x3C: /* HSW */
case 0x45: /* HSW */
case 0x46: /* HSW */
do_gfx_perf_limit_reasons = 1;
case 0x3F: /* HSX */
do_core_perf_limit_reasons = 1;
do_ring_perf_limit_reasons = 1;
default:
return;
}
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
int print_thermal(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
unsigned int dts;
int cpu;
if (!(do_dts || do_ptm))
return 0;
cpu = t->cpu_id;
/* DTS is per-core, no need to print for each thread */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE))
return 0;
if (cpu_migrate(cpu)) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "Could not migrate to CPU %d\n", cpu);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return -1;
}
if (do_ptm && (t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE)) {
if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_STATUS, &msr))
return 0;
dts = (msr >> 16) & 0x7F;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_IA32_PACKAGE_THERM_STATUS: 0x%08llx (%d C)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, tcc_activation_temp - dts);
#ifdef THERM_DEBUG
if (get_msr(cpu, MSR_IA32_PACKAGE_THERM_INTERRUPT, &msr))
return 0;
dts = (msr >> 16) & 0x7F;
dts2 = (msr >> 8) & 0x7F;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_IA32_PACKAGE_THERM_INTERRUPT: 0x%08llx (%d C, %d C)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, tcc_activation_temp - dts, tcc_activation_temp - dts2);
#endif
}
if (do_dts) {
unsigned int resolution;
if (get_msr(cpu, MSR_IA32_THERM_STATUS, &msr))
return 0;
dts = (msr >> 16) & 0x7F;
resolution = (msr >> 27) & 0xF;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_IA32_THERM_STATUS: 0x%08llx (%d C +/- %d)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, tcc_activation_temp - dts, resolution);
#ifdef THERM_DEBUG
if (get_msr(cpu, MSR_IA32_THERM_INTERRUPT, &msr))
return 0;
dts = (msr >> 16) & 0x7F;
dts2 = (msr >> 8) & 0x7F;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_IA32_THERM_INTERRUPT: 0x%08llx (%d C, %d C)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, tcc_activation_temp - dts, tcc_activation_temp - dts2);
#endif
}
return 0;
}
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
void print_power_limit_msr(int cpu, unsigned long long msr, char *label)
{
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: %s: %sabled (%f Watts, %f sec, clamp %sabled)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, label,
((msr >> 15) & 1) ? "EN" : "DIS",
((msr >> 0) & 0x7FFF) * rapl_power_units,
(1.0 + (((msr >> 22) & 0x3)/4.0)) * (1 << ((msr >> 17) & 0x1F)) * rapl_time_units,
(((msr >> 16) & 1) ? "EN" : "DIS"));
return;
}
int print_rapl(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
int cpu;
if (!do_rapl)
return 0;
/* RAPL counters are per package, so print only for 1st thread/package */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
return 0;
cpu = t->cpu_id;
if (cpu_migrate(cpu)) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "Could not migrate to CPU %d\n", cpu);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return -1;
}
if (get_msr(cpu, MSR_RAPL_POWER_UNIT, &msr))
return -1;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_RAPL_POWER_UNIT: 0x%08llx "
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
"(%f Watts, %f Joules, %f sec.)\n", cpu, msr,
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
rapl_power_units, rapl_energy_units, rapl_time_units);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
}
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
if (do_rapl & RAPL_PKG_POWER_INFO) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, MSR_PKG_POWER_INFO, &msr))
return -5;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_PKG_POWER_INFO: 0x%08llx (%.0f W TDP, RAPL %.0f - %.0f W, %f sec.)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr,
((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 16) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 32) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 48) & RAPL_TIME_GRANULARITY) * rapl_time_units);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
}
if (do_rapl & RAPL_PKG) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, MSR_PKG_POWER_LIMIT, &msr))
return -9;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_PKG_POWER_LIMIT: 0x%08llx (%slocked)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, (msr >> 63) & 1 ? "": "UN");
print_power_limit_msr(cpu, msr, "PKG Limit #1");
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: PKG Limit #2: %sabled (%f Watts, %f* sec, clamp %sabled)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu,
((msr >> 47) & 1) ? "EN" : "DIS",
((msr >> 32) & 0x7FFF) * rapl_power_units,
(1.0 + (((msr >> 54) & 0x3)/4.0)) * (1 << ((msr >> 49) & 0x1F)) * rapl_time_units,
((msr >> 48) & 1) ? "EN" : "DIS");
}
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
if (do_rapl & RAPL_DRAM_POWER_INFO) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, MSR_DRAM_POWER_INFO, &msr))
return -6;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_DRAM_POWER_INFO,: 0x%08llx (%.0f W TDP, RAPL %.0f - %.0f W, %f sec.)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr,
((msr >> 0) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 16) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 32) & RAPL_POWER_GRANULARITY) * rapl_power_units,
((msr >> 48) & RAPL_TIME_GRANULARITY) * rapl_time_units);
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
}
if (do_rapl & RAPL_DRAM) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, MSR_DRAM_POWER_LIMIT, &msr))
return -9;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_DRAM_POWER_LIMIT: 0x%08llx (%slocked)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, (msr >> 31) & 1 ? "": "UN");
print_power_limit_msr(cpu, msr, "DRAM Limit");
}
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
if (do_rapl & RAPL_CORE_POLICY) {
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, MSR_PP0_POLICY, &msr))
return -7;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_PP0_POLICY: %lld\n", cpu, msr & 0xF);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
}
}
if (do_rapl & RAPL_CORES) {
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, MSR_PP0_POWER_LIMIT, &msr))
return -9;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_PP0_POWER_LIMIT: 0x%08llx (%slocked)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, (msr >> 31) & 1 ? "": "UN");
print_power_limit_msr(cpu, msr, "Cores Limit");
}
}
if (do_rapl & RAPL_GFX) {
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug) {
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, MSR_PP1_POLICY, &msr))
return -8;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_PP1_POLICY: %lld\n", cpu, msr & 0xF);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
if (get_msr(cpu, MSR_PP1_POWER_LIMIT, &msr))
return -9;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_PP1_POWER_LIMIT: 0x%08llx (%slocked)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, (msr >> 31) & 1 ? "": "UN");
print_power_limit_msr(cpu, msr, "GFX Limit");
}
}
return 0;
}
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
/*
* SNB adds support for additional MSRs:
*
* MSR_PKG_C7_RESIDENCY 0x000003fa
* MSR_CORE_C7_RESIDENCY 0x000003fe
* MSR_PKG_C2_RESIDENCY 0x0000060d
*/
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
int has_snb_msrs(unsigned int family, unsigned int model)
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
if (!genuine_intel)
return 0;
switch (model) {
case 0x2A:
case 0x2D:
tools/power turbostat: fix IVB support Initial IVB support went into turbostat in Linux-3.1: 553575f1ae048aa44682b46b3c51929a0b3ad337 (tools turbostat: recognize and run properly on IVB) However, when running on IVB, turbostat would fail to report the new couters added with SNB, c7, pc2 and pc7. So in scenarios where these counters are non-zero on IVB, turbostat would report erroneous residencey results. In particular c7 time would be added to c1 time, since c1 time is calculated as "that which is left over". Also, turbostat reports MHz capabilities when passed the "-v" option, and it would incorrectly report 133MHz bclk instead of 100MHz bclk for IVB, which would inflate GHz reported with that option. This patch is a backport of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 11:34:44 +08:00
case 0x3A: /* IVB */
tools/power turbostat: run on IVB Xeon This fix is required to run on IVB Xeon, which previously had an incorrect cpuid model number listed. Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-27 06:11:31 +08:00
case 0x3E: /* IVB Xeon */
tools/power turbostat: support Haswell This patch enables turbostat to run properly on the next-generation Intel(R) Microarchitecture, code named "Haswell" (HSW). HSW supports the BCLK and counters found in SNB. Signed-off-by: Len Brown <len.brown@intel.com>
2013-01-08 14:26:07 +08:00
case 0x3C: /* HSW */
case 0x3F: /* HSW */
case 0x45: /* HSW */
tools/power turbostat: additional Haswell CPU-id There is an additional HSW CPU-id, 0x46, which has C-states exactly like CPU-id 0x45. Signed-off-by: Len Brown <len.brown@intel.com>
2013-03-15 22:58:02 +08:00
case 0x46: /* HSW */
tools/power turbostat: Run on Broadwell Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-28 12:28:53 +08:00
case 0x3D: /* BDW */
tools/power turbostat: support additional Broadwell model Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-11 04:38:04 +08:00
case 0x47: /* BDW */
tools/power turbostat: Run on Broadwell Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-28 12:28:53 +08:00
case 0x4F: /* BDX */
case 0x56: /* BDX-DE */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
case 0x4E: /* SKL */
case 0x5E: /* SKL */
tools/power turbostat: initial KBL support KBL is similar to SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:59 +08:00
case 0x8E: /* KBL */
case 0x9E: /* KBL */
tools/power turbostat: initial SKX support SKX has a lot in common with HSX Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:58 +08:00
case 0x55: /* SKX */
tools/power turbostat: initial BXT support Broxton has a lot in common with SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:56 +08:00
case 0x5C: /* BXT */
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
return 1;
}
return 0;
}
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
/*
* HSW adds support for additional MSRs:
*
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:55 +08:00
* MSR_PKG_C8_RESIDENCY 0x00000630
* MSR_PKG_C9_RESIDENCY 0x00000631
* MSR_PKG_C10_RESIDENCY 0x00000632
*
* MSR_PKGC8_IRTL 0x00000633
* MSR_PKGC9_IRTL 0x00000634
* MSR_PKGC10_IRTL 0x00000635
*
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
*/
int has_hsw_msrs(unsigned int family, unsigned int model)
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
{
if (!genuine_intel)
return 0;
switch (model) {
tools/power turbostat: Run on Broadwell Signed-off-by: Len Brown <len.brown@intel.com>
2014-02-28 12:28:53 +08:00
case 0x45: /* HSW */
case 0x3D: /* BDW */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
case 0x4E: /* SKL */
case 0x5E: /* SKL */
tools/power turbostat: initial KBL support KBL is similar to SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:59 +08:00
case 0x8E: /* KBL */
case 0x9E: /* KBL */
tools/power turbostat: initial BXT support Broxton has a lot in common with SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:56 +08:00
case 0x5C: /* BXT */
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
return 1;
}
return 0;
}
/*
* SKL adds support for additional MSRS:
*
* MSR_PKG_WEIGHTED_CORE_C0_RES 0x00000658
* MSR_PKG_ANY_CORE_C0_RES 0x00000659
* MSR_PKG_ANY_GFXE_C0_RES 0x0000065A
* MSR_PKG_BOTH_CORE_GFXE_C0_RES 0x0000065B
*/
int has_skl_msrs(unsigned int family, unsigned int model)
{
if (!genuine_intel)
return 0;
switch (model) {
case 0x4E: /* SKL */
case 0x5E: /* SKL */
tools/power turbostat: initial KBL support KBL is similar to SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:59 +08:00
case 0x8E: /* KBL */
case 0x9E: /* KBL */
tools/power turbostat: display C8, C9, C10 residency Display residency in the new C-states, C8, C9, C10. C8, C9, C10 are present on some: "Fourth Generation Intel(R) Core(TM) Processors", which are based on Intel(R) microarchitecture code name Haswell. Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-21 21:22:43 +08:00
return 1;
}
return 0;
}
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
int is_slm(unsigned int family, unsigned int model)
{
if (!genuine_intel)
return 0;
switch (model) {
case 0x37: /* BYT */
case 0x4D: /* AVN */
return 1;
}
return 0;
}
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
int is_knl(unsigned int family, unsigned int model)
{
if (!genuine_intel)
return 0;
switch (model) {
case 0x57: /* KNL */
return 1;
}
return 0;
}
tools/power turbostat: KNL workaround for %Busy and Avg_MHz KNL increments APERF and MPERF every 1024 clocks. This is compliant with the architecture specification, which requires that only the ratio of APERF/MPERF need be valid. However, turbostat takes advantage of the fact that these two MSRs increment every un-halted clock at the actual and base frequency: AVG_MHz = APERF_delta/measurement_interval %Busy = MPERF_delta/TSC_delta This quirk is needed for these calculations to also work on KNL, which would otherwise show a value 1024x smaller than expected. Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-14 19:31:00 +08:00
unsigned int get_aperf_mperf_multiplier(unsigned int family, unsigned int model)
{
if (is_knl(family, model))
return 1024;
return 1;
}
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
#define SLM_BCLK_FREQS 5
double slm_freq_table[SLM_BCLK_FREQS] = { 83.3, 100.0, 133.3, 116.7, 80.0};
double slm_bclk(void)
{
unsigned long long msr = 3;
unsigned int i;
double freq;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
if (get_msr(base_cpu, MSR_FSB_FREQ, &msr))
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "SLM BCLK: unknown\n");
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
i = msr & 0xf;
if (i >= SLM_BCLK_FREQS) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "SLM BCLK[%d] invalid\n", i);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
msr = 3;
}
freq = slm_freq_table[i];
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "SLM BCLK: %.1f Mhz\n", freq);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
return freq;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
double discover_bclk(unsigned int family, unsigned int model)
{
tools/power turbostat: Intel Xeon x200: fix erroneous bclk value x200 does not enable any way to programmatically obtain bus clock speed. Bclk for the architecture has a fixed value of 100 MHz. At the same time x200 cannot be included in has_snb_msrs since it does not support C7 idle state. prior to this patch, MHz values reported on this chip were erroneously calculated using bclk of 133MHz, causing MHz values to be reported 33% higher than actual. Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-10 23:35:17 +08:00
if (has_snb_msrs(family, model) || is_knl(family, model))
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
return 100.00;
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
else if (is_slm(family, model))
return slm_bclk();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
else
return 133.33;
}
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
/*
* MSR_IA32_TEMPERATURE_TARGET indicates the temperature where
* the Thermal Control Circuit (TCC) activates.
* This is usually equal to tjMax.
*
* Older processors do not have this MSR, so there we guess,
* but also allow cmdline over-ride with -T.
*
* Several MSR temperature values are in units of degrees-C
* below this value, including the Digital Thermal Sensor (DTS),
* Package Thermal Management Sensor (PTM), and thermal event thresholds.
*/
int set_temperature_target(struct thread_data *t, struct core_data *c, struct pkg_data *p)
{
unsigned long long msr;
unsigned int target_c_local;
int cpu;
/* tcc_activation_temp is used only for dts or ptm */
if (!(do_dts || do_ptm))
return 0;
/* this is a per-package concept */
if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE) || !(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
return 0;
cpu = t->cpu_id;
if (cpu_migrate(cpu)) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "Could not migrate to CPU %d\n", cpu);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return -1;
}
if (tcc_activation_temp_override != 0) {
tcc_activation_temp = tcc_activation_temp_override;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: Using cmdline TCC Target (%d C)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, tcc_activation_temp);
return 0;
}
/* Temperature Target MSR is Nehalem and newer only */
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
if (!do_nhm_platform_info)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
goto guess;
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
if (get_msr(base_cpu, MSR_IA32_TEMPERATURE_TARGET, &msr))
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
goto guess;
tools / power: turbostat: Drop temperature checks The Intel 64 and IA-32 Architectures Software Developer's Manual says that TjMax is stored in bits 23:16 of MSR_TEMPERATURE TARGET (0x1a2). That's 8 bits, not 7, so it must be masked with 0xFF rather than 0x7F. The manual has no mention of which values should be considered valid, which kind of implies that they all are. Arbitrarily discarding values outside a specific range is wrong. The upper range check had to be fixed recently (commit 144b44b1) and the lower range check is just as wrong. See bug #75071: https://bugzilla.kernel.org/show_bug.cgi?id=75071 There are many Xeon processor series with TjMax of 70, 71 or 80 degrees Celsius, way below the arbitrary 85 degrees Celsius limit. There may be other (past or future) models with even lower limits. So drop this arbitrary check. The only value that would be clearly invalid is 0. Everything else should be accepted. After these changes, turbostat is aligned with what the coretemp driver does. Signed-off-by: Jean Delvare <jdelvare@suse.de> Cc: Len Brown <len.brown@intel.com> Acked-by: Guenter Roeck <linux@roeck-us.net> Reviewed-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-05-01 17:40:19 +08:00
target_c_local = (msr >> 16) & 0xFF;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_IA32_TEMPERATURE_TARGET: 0x%08llx (%d C)\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, msr, target_c_local);
tools / power: turbostat: Drop temperature checks The Intel 64 and IA-32 Architectures Software Developer's Manual says that TjMax is stored in bits 23:16 of MSR_TEMPERATURE TARGET (0x1a2). That's 8 bits, not 7, so it must be masked with 0xFF rather than 0x7F. The manual has no mention of which values should be considered valid, which kind of implies that they all are. Arbitrarily discarding values outside a specific range is wrong. The upper range check had to be fixed recently (commit 144b44b1) and the lower range check is just as wrong. See bug #75071: https://bugzilla.kernel.org/show_bug.cgi?id=75071 There are many Xeon processor series with TjMax of 70, 71 or 80 degrees Celsius, way below the arbitrary 85 degrees Celsius limit. There may be other (past or future) models with even lower limits. So drop this arbitrary check. The only value that would be clearly invalid is 0. Everything else should be accepted. After these changes, turbostat is aligned with what the coretemp driver does. Signed-off-by: Jean Delvare <jdelvare@suse.de> Cc: Len Brown <len.brown@intel.com> Acked-by: Guenter Roeck <linux@roeck-us.net> Reviewed-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-05-01 17:40:19 +08:00
if (!target_c_local)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
goto guess;
tcc_activation_temp = target_c_local;
return 0;
guess:
tcc_activation_temp = TJMAX_DEFAULT;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: Guessing tjMax %d C, Please use -T to specify\n",
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu, tcc_activation_temp);
return 0;
}
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-22 01:22:47 +08:00
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-12 02:26:03 +08:00
void decode_feature_control_msr(void)
{
unsigned long long msr;
if (!get_msr(base_cpu, MSR_IA32_FEATURE_CONTROL, &msr))
fprintf(outf, "cpu%d: MSR_IA32_FEATURE_CONTROL: 0x%08llx (%sLocked %s)\n",
base_cpu, msr,
msr & FEATURE_CONTROL_LOCKED ? "" : "UN-",
msr & (1 << 18) ? "SGX" : "");
}
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-22 01:22:47 +08:00
void decode_misc_enable_msr(void)
{
unsigned long long msr;
if (!get_msr(base_cpu, MSR_IA32_MISC_ENABLE, &msr))
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_IA32_MISC_ENABLE: 0x%08llx (%s %s %s)\n",
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-22 01:22:47 +08:00
base_cpu, msr,
msr & (1 << 3) ? "TCC" : "",
msr & (1 << 16) ? "EIST" : "",
msr & (1 << 18) ? "MONITOR" : "");
}
tools/power turbostat: Decode MSR_MISC_PWR_MGMT This MSR is helpful to show if P-state HW coordination is enabled or disabled. Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-03 14:35:36 +08:00
/*
* Decode MSR_MISC_PWR_MGMT
*
* Decode the bits according to the Nehalem documentation
* bit[0] seems to continue to have same meaning going forward
* bit[1] less so...
*/
void decode_misc_pwr_mgmt_msr(void)
{
unsigned long long msr;
if (!do_nhm_platform_info)
return;
if (!get_msr(base_cpu, MSR_MISC_PWR_MGMT, &msr))
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d: MSR_MISC_PWR_MGMT: 0x%08llx (%sable-EIST_Coordination %sable-EPB)\n",
tools/power turbostat: Decode MSR_MISC_PWR_MGMT This MSR is helpful to show if P-state HW coordination is enabled or disabled. Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-03 14:35:36 +08:00
base_cpu, msr,
msr & (1 << 0) ? "DIS" : "EN",
msr & (1 << 1) ? "EN" : "DIS");
}
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
void process_cpuid()
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 15:30:51 +08:00
unsigned int eax, ebx, ecx, edx, max_level, max_extended_level;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
unsigned int fms, family, model, stepping;
eax = ebx = ecx = edx = 0;
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 15:14:35 +08:00
__cpuid(0, max_level, ebx, ecx, edx);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
if (ebx == 0x756e6547 && edx == 0x49656e69 && ecx == 0x6c65746e)
genuine_intel = 1;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "CPUID(0): %.4s%.4s%.4s ",
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
(char *)&ebx, (char *)&edx, (char *)&ecx);
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 15:14:35 +08:00
__cpuid(1, fms, ebx, ecx, edx);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
family = (fms >> 8) & 0xf;
model = (fms >> 4) & 0xf;
stepping = fms & 0xf;
if (family == 6 || family == 0xf)
model += ((fms >> 16) & 0xf) << 4;
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-22 01:22:47 +08:00
if (debug) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%d CPUID levels; family:model:stepping 0x%x:%x:%x (%d:%d:%d)\n",
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
max_level, family, model, stepping, family, model, stepping);
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-12 02:26:03 +08:00
fprintf(outf, "CPUID(1): %s %s %s %s %s %s %s %s %s\n",
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-22 01:22:47 +08:00
ecx & (1 << 0) ? "SSE3" : "-",
ecx & (1 << 3) ? "MONITOR" : "-",
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-12 02:26:03 +08:00
ecx & (1 << 6) ? "SMX" : "-",
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-22 01:22:47 +08:00
ecx & (1 << 7) ? "EIST" : "-",
ecx & (1 << 8) ? "TM2" : "-",
edx & (1 << 4) ? "TSC" : "-",
edx & (1 << 5) ? "MSR" : "-",
edx & (1 << 22) ? "ACPI-TM" : "-",
edx & (1 << 29) ? "TM" : "-");
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (!(edx & (1 << 5)))
errx(1, "CPUID: no MSR");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
/*
* check max extended function levels of CPUID.
* This is needed to check for invariant TSC.
* This check is valid for both Intel and AMD.
*/
ebx = ecx = edx = 0;
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 15:14:35 +08:00
__cpuid(0x80000000, max_extended_level, ebx, ecx, edx);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 15:30:51 +08:00
if (max_extended_level >= 0x80000007) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
/*
* Non-Stop TSC is advertised by CPUID.EAX=0x80000007: EDX.bit8
* this check is valid for both Intel and AMD
*/
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 15:14:35 +08:00
__cpuid(0x80000007, eax, ebx, ecx, edx);
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
has_invariant_tsc = edx & (1 << 8);
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
/*
* APERF/MPERF is advertised by CPUID.EAX=0x6: ECX.bit0
* this check is valid for both Intel and AMD
*/
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 15:14:35 +08:00
__cpuid(0x6, eax, ebx, ecx, edx);
tools: turbostat: fix bitwise and operand bug could cause false positive on indicating presence of invarient TSC or APERF support. Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Len Brown <len.brown@intel.com>
2011-01-21 22:11:19 +08:00
has_aperf = ecx & (1 << 0);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
do_dts = eax & (1 << 0);
do_ptm = eax & (1 << 6);
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
has_hwp = eax & (1 << 7);
has_hwp_notify = eax & (1 << 8);
has_hwp_activity_window = eax & (1 << 9);
has_hwp_epp = eax & (1 << 10);
has_hwp_pkg = eax & (1 << 11);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
has_epb = ecx & (1 << 3);
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "CPUID(6): %sAPERF, %sDTS, %sPTM, %sHWP, "
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
"%sHWPnotify, %sHWPwindow, %sHWPepp, %sHWPpkg, %sEPB\n",
has_aperf ? "" : "No-",
do_dts ? "" : "No-",
do_ptm ? "" : "No-",
has_hwp ? "" : "No-",
has_hwp_notify ? "" : "No-",
has_hwp_activity_window ? "" : "No-",
has_hwp_epp ? "" : "No-",
has_hwp_pkg ? "" : "No-",
has_epb ? "" : "No-");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: decode more CPUID fields for debugging, dump a few more fields: CPUID(1): SSE3 MONITOR EIST TM2 TSC MSR ACPI-TM TM cpu0: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MONITOR) Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-22 01:22:47 +08:00
if (debug)
decode_misc_enable_msr();
tools/power turbostat: SGX state should print only if --debug The CPUID.SGX bit was printed, even if --debug was used Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:54 +08:00
if (max_level >= 0x7 && debug) {
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-12 02:26:03 +08:00
int has_sgx;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: indicate SMX and SGX support SGX presence is related to a SKL power workaround, so lets show when that is enabled. Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-12 02:26:03 +08:00
ecx = 0;
__cpuid_count(0x7, 0, eax, ebx, ecx, edx);
has_sgx = ebx & (1 << 2);
fprintf(outf, "CPUID(7): %sSGX\n", has_sgx ? "" : "No-");
if (has_sgx)
decode_feature_control_msr();
}
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 15:30:51 +08:00
if (max_level >= 0x15) {
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
unsigned int eax_crystal;
unsigned int ebx_tsc;
/*
* CPUID 15H TSC/Crystal ratio, possibly Crystal Hz
*/
eax_crystal = ebx_tsc = crystal_hz = edx = 0;
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 15:14:35 +08:00
__cpuid(0x15, eax_crystal, ebx_tsc, crystal_hz, edx);
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
if (ebx_tsc != 0) {
if (debug && (ebx != 0))
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "CPUID(0x15): eax_crystal: %d ebx_tsc: %d ecx_crystal_hz: %d\n",
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
eax_crystal, ebx_tsc, crystal_hz);
if (crystal_hz == 0)
switch(model) {
case 0x4E: /* SKL */
case 0x5E: /* SKL */
tools/power turbostat: initial KBL support KBL is similar to SKL Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:59 +08:00
case 0x8E: /* KBL */
case 0x9E: /* KBL */
tools/power turbostat: decode BXT TSC frequency via CPUID Hard-code BXT ART to 19200MHz, so turbostat --debug can fully enumerate TSC: CPUID(0x15): eax_crystal: 3 ebx_tsc: 186 ecx_crystal_hz: 0 TSC: 1190 MHz (19200000 Hz * 186 / 3 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:57 +08:00
crystal_hz = 24000000; /* 24.0 MHz */
break;
tools/power turbostat: initial SKX support SKX has a lot in common with HSX Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:58 +08:00
case 0x55: /* SKX */
crystal_hz = 25000000; /* 25.0 MHz */
break;
tools/power turbostat: decode BXT TSC frequency via CPUID Hard-code BXT ART to 19200MHz, so turbostat --debug can fully enumerate TSC: CPUID(0x15): eax_crystal: 3 ebx_tsc: 186 ecx_crystal_hz: 0 TSC: 1190 MHz (19200000 Hz * 186 / 3 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:57 +08:00
case 0x5C: /* BXT */
crystal_hz = 19200000; /* 19.2 MHz */
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
break;
default:
crystal_hz = 0;
}
if (crystal_hz) {
tsc_hz = (unsigned long long) crystal_hz * ebx_tsc / eax_crystal;
if (debug)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "TSC: %lld MHz (%d Hz * %d / %d / 1000000)\n",
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
tsc_hz / 1000000, crystal_hz, ebx_tsc, eax_crystal);
}
}
}
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 15:30:51 +08:00
if (max_level >= 0x16) {
unsigned int base_mhz, max_mhz, bus_mhz, edx;
/*
* CPUID 16H Base MHz, Max MHz, Bus MHz
*/
base_mhz = max_mhz = bus_mhz = edx = 0;
tools/power turbostat: call __cpuid() instead of __get_cpuid() turbostat already checks whether calling each cpuid leavf is legal, and it doesn't look at the function return value, so call the simpler gcc intrinsic __cpuid() instead of __get_cpuid(). syntax only, no functional change Signed-off-by: Len Brown <len.brown@intel.com>
2016-03-13 15:14:35 +08:00
__cpuid(0x16, base_mhz, max_mhz, bus_mhz, edx);
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 15:30:51 +08:00
if (debug)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "CPUID(0x16): base_mhz: %d max_mhz: %d bus_mhz: %d\n",
tools/power turbostat: CPUID(0x16) leaf shows base, max, and bus frequency This CPUID leaf is available on Skylake: CPUID(0x16): base_mhz: 1500 max_mhz: 2200 bus_mhz: 100 Signed-off-by: Len Brown <len.brown@intel.com>
2015-11-23 15:30:51 +08:00
base_mhz, max_mhz, bus_mhz);
}
tools/power turbostat: calculate TSC frequency from CPUID(0x15) on SKL turbostat --debug ... CPUID(0x15): eax_crystal: 2 ebx_tsc: 100 ecx_crystal_hz: 0 TSC: 1200 MHz (24000000 Hz * 100 / 2 / 1000000) Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-02 09:02:57 +08:00
tools/power turbostat: KNL workaround for %Busy and Avg_MHz KNL increments APERF and MPERF every 1024 clocks. This is compliant with the architecture specification, which requires that only the ratio of APERF/MPERF need be valid. However, turbostat takes advantage of the fact that these two MSRs increment every un-halted clock at the actual and base frequency: AVG_MHz = APERF_delta/measurement_interval %Busy = MPERF_delta/TSC_delta This quirk is needed for these calculations to also work on KNL, which would otherwise show a value 1024x smaller than expected. Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-14 19:31:00 +08:00
if (has_aperf)
aperf_mperf_multiplier = get_aperf_mperf_multiplier(family, model);
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
do_nhm_platform_info = do_nhm_cstates = do_smi = probe_nhm_msrs(family, model);
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
do_snb_cstates = has_snb_msrs(family, model);
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:55 +08:00
do_irtl_snb = has_snb_msrs(family, model);
tools/power turbostat: Skip printing disabled package C-states Replaced previously open-coded Package C-state Limit decoding with table-driven decoding. In doing so, updated to match January 2015 "Intel(R) 64 and IA-23 Architectures Software Developer's Manual". In the past, turbostat would print package C-state residency columns for all package states supported by the model's architecture, even though a particular SKU may not support them, or they may be disabled by the BIOS. Now turbostat will skip printing colunns if MSRs indicate that they are not enabled. eg. many SKUs don't support PC7, and so that column will no longer be printed. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 12:39:45 +08:00
do_pc2 = do_snb_cstates && (pkg_cstate_limit >= PCL__2);
do_pc3 = (pkg_cstate_limit >= PCL__3);
do_pc6 = (pkg_cstate_limit >= PCL__6);
do_pc7 = do_snb_cstates && (pkg_cstate_limit >= PCL__7);
tools/power turbostat: relax dependency on invariant TSC Turbostat can be useful on systems that do not support invariant TSC, so allow it to run on those systgems. All arithmetic in turbostat using the TSC value is per-processsor, so it does not depend on the TSC values being in sync acrosss processors. Turbostat uses gettimeofday() for the measurement interval rather than using the TSC directly, so that key metric is also immune from variable TSC. Turbostat prints a TSC sanity check column: TSC_MHz = TSC_delta/interval If this column is constant and is close to the processor base frequency, then the TSC is behaving properly. The other key turbostat columns are calculated this way: Avg_Mhz = APERF_delta/interval %Busy = MPERF_delta/TSC_delta Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/interval Tested on Core2 and Core2-Xeon, and so this patch includes a few other changes to remove the assumption that target systems are Nehalem and newer. Signed-off-by: Len Brown <len.brown@intel.com>
2015-01-23 13:12:33 +08:00
do_c8_c9_c10 = has_hsw_msrs(family, model);
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:55 +08:00
do_irtl_hsw = has_hsw_msrs(family, model);
tools/power turbostat: Initial Skylake support Skylake adds some additional residency counters. Skylake supports a different mix of RAPL registers from any previous product. In most other ways, Skylake is like Broadwell. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-26 12:50:30 +08:00
do_skl_residency = has_skl_msrs(family, model);
tools / power turbostat: Support Silvermont Support the next generation Intel Atom processor mirco-architecture, formerly called Silvermont. The server version, formerly called "Avoton", is named the "Intel(R) Atom(TM) Processor C2000 Product Family". The client version, formerly called "Bay Trail", is named the "Intel Atom Processor Z3000 Series", as well as various "Intel Pentium Processor" and "Intel Celeron Processor" brands, depending on form-factor. Silvermont has a set of MSRs not far off from NHM, but the RAPL register set is a sub-set of those previously supported. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-11-09 13:30:16 +08:00
do_slm_cstates = is_slm(family, model);
tools/power turbostat: enable turbostat to support Knights Landing (KNL) Changes mainly to account for minor differences in Knights Landing(KNL): 1. KNL supports C1 and C6 core states. 2. KNL supports PC2, PC3 and PC6 package states. 3. KNL has a different encoding of the TURBO_RATIO_LIMIT MSR Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-21 00:49:34 +08:00
do_knl_cstates = is_knl(family, model);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: Decode MSR_MISC_PWR_MGMT This MSR is helpful to show if P-state HW coordination is enabled or disabled. Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-03 14:35:36 +08:00
if (debug)
decode_misc_pwr_mgmt_msr();
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
rapl_probe(family, model);
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
perf_limit_reasons_probe(family, model);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
if (debug)
tools/power turbostat: fix various build warnings When building with gcc 6 we're getting various build warnings that just require some trivial function declaration and call fixes: turbostat.c: In function ‘dump_cstate_pstate_config_info’: turbostat.c:1973:1: warning: type of ‘family’ defaults to ‘int’ dump_cstate_pstate_config_info(family, model) turbostat.c:1973:1: warning: type of ‘model’ defaults to ‘int’ turbostat.c: In function ‘get_tdp’: turbostat.c:2145:8: warning: type of ‘model’ defaults to ‘int’ double get_tdp(model) turbostat.c: In function ‘perf_limit_reasons_probe’: turbostat.c:2259:6: warning: type of ‘family’ defaults to ‘int’ void perf_limit_reasons_probe(family, model) turbostat.c:2259:6: warning: type of ‘model’ defaults to ‘int’ Signed-off-by: Colin Ian King <colin.king@canonical.com> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/n/tip-wbicer8n0s9qe6ql8h9x478e@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-03-02 21:50:25 +08:00
dump_cstate_pstate_config_info(family, model);
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
tools/power turbostat: SKL: Adjust for TSC difference from base frequency On a Skylake with 1500MHz base frequency, the TSC runs at 1512MHz. This is because the TSC is no longer in the n*100 MHz BCLK domain, but is now in the m*24MHz crystal clock domain. (24 MHz * 63 = 1512 MHz) This adds error to several calculations in turbostat, unless the TSC sample sizes are adjusted for this difference. Note that calculations in the time domain are immune from this issue, as the timing sub-system has already calibrated the TSC against a known wall clock. AVG_MHz = APERF_delta/measurement_interval need no adjustment. APERF_delta is in the BCLK domain, and measurement_interval is in the time domain. TSC_MHz = TSC_delta/measurement_interval needs no adjustment -- as we really do want to report the actual measured TSC delta here, and measurement_interval is in the accurate time domain. %Busy = MPERF_delta/TSC_delta needs adjustment to use TSC_BCLK_DOMAIN_delta. TSC_BCLK_DOMAIN_delta = TSC_delta * base_hz / tsc_hz Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval need adjustment as above. No other metrics in turbostat need to be adjusted. Before: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 24.84 2216 1512 0 2191 98.73 2219 1514 2 0 0.01 2130 1512 1 9 0.43 2016 1512 3 2 0.08 2016 1512 After: CPU Avg_MHz %Busy Bzy_MHz TSC_MHz - 550 25.05 2198 1512 0 2190 99.62 2199 1512 2 0 0.01 2152 1512 1 9 0.46 2000 1512 3 2 0.10 2000 1512 Note that in this example, the "Before" Bzy_MHz was reported as exceeding the 2200 max turbo rate. Also, even a pinned spin loop would not be reported as over 99% busy. Signed-off-by: Len Brown <len.brown@intel.com>
2015-09-26 12:12:38 +08:00
if (has_skl_msrs(family, model))
calculate_tsc_tweak();
tools/power turbostat: show GFX%rc6 The column "GFX%c6" show the percentage of time the GPU is in the "render C6" state, rc6. Deep package C-states on several systems depend on the GPU being in RC6. This information comes from the counter /sys/class/drm/card0/power/rc6_residency_ms, as read before and after the measurement interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 14:28:12 +08:00
do_gfx_rc6_ms = !access("/sys/class/drm/card0/power/rc6_residency_ms", R_OK);
tools/power turbostat: show GFXMHz Under the column "GFXMHz", show a snapshot of this attribute: /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz This is an instantaneous snapshot of what sysfs presents at the end of the measurement interval. turbostat does not average or otherwise perform any math on this value. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 13:37:54 +08:00
do_gfx_mhz = !access("/sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz", R_OK);
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
return;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
void help()
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
{
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf,
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
"Usage: turbostat [OPTIONS][(--interval seconds) | COMMAND ...]\n"
"\n"
"Turbostat forks the specified COMMAND and prints statistics\n"
"when COMMAND completes.\n"
"If no COMMAND is specified, turbostat wakes every 5-seconds\n"
"to print statistics, until interrupted.\n"
"--debug run in \"debug\" mode\n"
"--interval sec Override default 5-second measurement interval\n"
"--help print this help message\n"
"--counter msr print 32-bit counter at address \"msr\"\n"
"--Counter msr print 64-bit Counter at address \"msr\"\n"
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
"--out file create or truncate \"file\" for all output\n"
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
"--msr msr print 32-bit value at address \"msr\"\n"
"--MSR msr print 64-bit Value at address \"msr\"\n"
"--version print version information\n"
"\n"
"For more help, run \"man turbostat\"\n");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
/*
* in /dev/cpu/ return success for names that are numbers
* ie. filter out ".", "..", "microcode".
*/
int dir_filter(const struct dirent *dirp)
{
if (isdigit(dirp->d_name[0]))
return 1;
else
return 0;
}
int open_dev_cpu_msr(int dummy1)
{
return 0;
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
void topology_probe()
{
int i;
int max_core_id = 0;
int max_package_id = 0;
int max_siblings = 0;
struct cpu_topology {
int core_id;
int physical_package_id;
} *cpus;
/* Initialize num_cpus, max_cpu_num */
topo.num_cpus = 0;
topo.max_cpu_num = 0;
for_all_proc_cpus(count_cpus);
if (!summary_only && topo.num_cpus > 1)
show_cpu = 1;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "num_cpus %d max_cpu_num %d\n", topo.num_cpus, topo.max_cpu_num);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
cpus = calloc(1, (topo.max_cpu_num + 1) * sizeof(struct cpu_topology));
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (cpus == NULL)
err(1, "calloc cpus");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/*
* Allocate and initialize cpu_present_set
*/
cpu_present_set = CPU_ALLOC((topo.max_cpu_num + 1));
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (cpu_present_set == NULL)
err(3, "CPU_ALLOC");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
cpu_present_setsize = CPU_ALLOC_SIZE((topo.max_cpu_num + 1));
CPU_ZERO_S(cpu_present_setsize, cpu_present_set);
for_all_proc_cpus(mark_cpu_present);
/*
* Allocate and initialize cpu_affinity_set
*/
cpu_affinity_set = CPU_ALLOC((topo.max_cpu_num + 1));
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (cpu_affinity_set == NULL)
err(3, "CPU_ALLOC");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
cpu_affinity_setsize = CPU_ALLOC_SIZE((topo.max_cpu_num + 1));
CPU_ZERO_S(cpu_affinity_setsize, cpu_affinity_set);
/*
* For online cpus
* find max_core_id, max_package_id
*/
for (i = 0; i <= topo.max_cpu_num; ++i) {
int siblings;
if (cpu_is_not_present(i)) {
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu%d NOT PRESENT\n", i);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
continue;
}
cpus[i].core_id = get_core_id(i);
if (cpus[i].core_id > max_core_id)
max_core_id = cpus[i].core_id;
cpus[i].physical_package_id = get_physical_package_id(i);
if (cpus[i].physical_package_id > max_package_id)
max_package_id = cpus[i].physical_package_id;
siblings = get_num_ht_siblings(i);
if (siblings > max_siblings)
max_siblings = siblings;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "cpu %d pkg %d core %d\n",
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
i, cpus[i].physical_package_id, cpus[i].core_id);
}
topo.num_cores_per_pkg = max_core_id + 1;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "max_core_id %d, sizing for %d cores per package\n",
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
max_core_id, topo.num_cores_per_pkg);
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 13:34:57 +08:00
if (debug && !summary_only && topo.num_cores_per_pkg > 1)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
show_core = 1;
topo.num_packages = max_package_id + 1;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "max_package_id %d, sizing for %d packages\n",
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
max_package_id, topo.num_packages);
tools/power turbostat: simplify default output Casual turbostat users generally just want to know MHz. So by default, just print enough information to make sense of MHz. All the other configuration data and columns for C-states and temperature etc, are printed with the --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-23 13:34:57 +08:00
if (debug && !summary_only && topo.num_packages > 1)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
show_pkg = 1;
topo.num_threads_per_core = max_siblings;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "max_siblings %d\n", max_siblings);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
free(cpus);
}
void
allocate_counters(struct thread_data **t, struct core_data **c, struct pkg_data **p)
{
int i;
*t = calloc(topo.num_threads_per_core * topo.num_cores_per_pkg *
topo.num_packages, sizeof(struct thread_data));
if (*t == NULL)
goto error;
for (i = 0; i < topo.num_threads_per_core *
topo.num_cores_per_pkg * topo.num_packages; i++)
(*t)[i].cpu_id = -1;
*c = calloc(topo.num_cores_per_pkg * topo.num_packages,
sizeof(struct core_data));
if (*c == NULL)
goto error;
for (i = 0; i < topo.num_cores_per_pkg * topo.num_packages; i++)
(*c)[i].core_id = -1;
*p = calloc(topo.num_packages, sizeof(struct pkg_data));
if (*p == NULL)
goto error;
for (i = 0; i < topo.num_packages; i++)
(*p)[i].package_id = i;
return;
error:
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
err(1, "calloc counters");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
/*
* init_counter()
*
* set cpu_id, core_num, pkg_num
* set FIRST_THREAD_IN_CORE and FIRST_CORE_IN_PACKAGE
*
* increment topo.num_cores when 1st core in pkg seen
*/
void init_counter(struct thread_data *thread_base, struct core_data *core_base,
struct pkg_data *pkg_base, int thread_num, int core_num,
int pkg_num, int cpu_id)
{
struct thread_data *t;
struct core_data *c;
struct pkg_data *p;
t = GET_THREAD(thread_base, thread_num, core_num, pkg_num);
c = GET_CORE(core_base, core_num, pkg_num);
p = GET_PKG(pkg_base, pkg_num);
t->cpu_id = cpu_id;
if (thread_num == 0) {
t->flags |= CPU_IS_FIRST_THREAD_IN_CORE;
if (cpu_is_first_core_in_package(cpu_id))
t->flags |= CPU_IS_FIRST_CORE_IN_PACKAGE;
}
c->core_id = core_num;
p->package_id = pkg_num;
}
int initialize_counters(int cpu_id)
{
int my_thread_id, my_core_id, my_package_id;
my_package_id = get_physical_package_id(cpu_id);
my_core_id = get_core_id(cpu_id);
tools/power turbostat: correctly display more than 2 threads/core Without this update, turbostat displays only 2 threads per core. Some processors, such as Xeon Phi, have more. Signed-off-by: Dasaratharaman Chandramouli <dasaratharaman.chandramouli@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-04-16 01:09:50 +08:00
my_thread_id = get_cpu_position_in_core(cpu_id);
if (!my_thread_id)
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
topo.num_cores++;
init_counter(EVEN_COUNTERS, my_thread_id, my_core_id, my_package_id, cpu_id);
init_counter(ODD_COUNTERS, my_thread_id, my_core_id, my_package_id, cpu_id);
return 0;
}
void allocate_output_buffer()
{
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
output_buffer = calloc(1, (1 + topo.num_cpus) * 1024);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
outp = output_buffer;
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (outp == NULL)
err(-1, "calloc output buffer");
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
}
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
void allocate_fd_percpu(void)
{
fd_percpu = calloc(topo.max_cpu_num, sizeof(int));
if (fd_percpu == NULL)
err(-1, "calloc fd_percpu");
}
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
void allocate_irq_buffers(void)
{
irq_column_2_cpu = calloc(topo.num_cpus, sizeof(int));
if (irq_column_2_cpu == NULL)
err(-1, "calloc %d", topo.num_cpus);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
irqs_per_cpu = calloc(topo.max_cpu_num, sizeof(int));
if (irqs_per_cpu == NULL)
err(-1, "calloc %d", topo.max_cpu_num);
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
void setup_all_buffers(void)
{
topology_probe();
tools/power turbostat: show IRQs per CPU The new IRQ column shows how many interrupts have occurred on each CPU during the measurement inteval. This information comes from the difference between /proc/interrupts shapshots made before and after the measurement interval. The first row, the system summary, shows the sum of the IRQS for all CPUs during that interval. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 12:48:05 +08:00
allocate_irq_buffers();
tools/power turbostat: make fewer systems calls skip the open(2)/close(2) on each msr read by keeping the /dev/cpu/*/msr files open. The remaining read(2) is generally far fewer cycles than the removed open(2) system call. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-27 09:51:02 +08:00
allocate_fd_percpu();
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
allocate_counters(&thread_even, &core_even, &package_even);
allocate_counters(&thread_odd, &core_odd, &package_odd);
allocate_output_buffer();
for_all_proc_cpus(initialize_counters);
}
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
void set_base_cpu(void)
{
base_cpu = sched_getcpu();
if (base_cpu < 0)
err(-ENODEV, "No valid cpus found");
if (debug > 1)
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "base_cpu = %d\n", base_cpu);
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
void turbostat_init()
{
tools/power turbostat: allow running without cpu0 Linux-3.7 added CONFIG_BOOTPARAM_HOTPLUG_CPU0, allowing systems to offline cpu0. But when cpu0 is offline, turbostat will not run: # turbostat ls turbostat: no /dev/cpu/0/msr This patch replaces the hard-coded use of cpu0 in turbostat with the current cpu, allowing it to run without a cpu0. Fewer cross-calls may also be needed due to use of current cpu, though this hard-coding was used only for the --debug preamble. Signed-off-by: Prarit Bhargava <prarit@redhat.com> Signed-off-by: Len Brown <len.brown@intel.com>
2015-05-25 20:34:28 +08:00
setup_all_buffers();
set_base_cpu();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
check_dev_msr();
tools/power turbostat: relax dependency on root permission For turbostat to run as non-root, it needs to permissions: 1. read access to /dev/cpu/*/msr via standard user/group/world file permissions 2. CAP_SYS_RAWIO eg. # setcap cap_sys_rawio=ep turbostat Yes, running as root still works. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 12:36:50 +08:00
check_permissions();
tools/power turbostat: dump MSR_TURBO_RATIO_LIMIT2 and up to 18 cores of turbo ratio limit when using the turbostat --debug option. Signed-off-by: Len Brown <len.brown@intel.com>
2015-03-24 08:29:09 +08:00
process_cpuid();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: decode HWP registers # turbostat --debug ... CPUID(6): ... HWP, HWPnotify, HWPwindow, HWPepp, HWPpkg ... ... cpu0: MSR_PM_ENABLE: 0x00000001 (HWP) cpu0: MSR_HWP_CAPABILITIES: 0x01050916 (high 0x16 guar 0x9 eff 0x5 low 0x1) cpu0: MSR_HWP_REQUEST: 0x80001604 (min 0x4 max 0x16 des 0x0 epp 0x80 window 0x0 pkg 0x0) cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min) cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min) Signed-off-by: Len Brown <len.brown@intel.com>
2015-12-01 14:36:39 +08:00
if (debug)
for_all_cpus(print_hwp, ODD_COUNTERS);
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
for_all_cpus(print_epb, ODD_COUNTERS);
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
tools/power turbostat: decode MSR_*_PERF_LIMIT_REASONS The Processor generation code-named Haswell added MSR_{CORE | GFX | RING}_PERF_LIMIT_REASONS to explain when and how the processor limits frequency. turbostat -v will now decode these bits. Each MSR has an "Active" set of bits which describe current conditions, and a "Logged" set of bits, which describe what has happened since last cleared. Turbostat currently doesn't clear the log bits. Signed-off-by: Len Brown <len.brown@intel.com>
2014-08-15 14:39:52 +08:00
for_all_cpus(print_perf_limit, ODD_COUNTERS);
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
for_all_cpus(print_rapl, ODD_COUNTERS);
for_all_cpus(set_temperature_target, ODD_COUNTERS);
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
for_all_cpus(print_thermal, ODD_COUNTERS);
tools/power turbostat: print IRTL MSRs Some processors use the Interrupt Response Time Limit (IRTL) MSR value to describe the maximum IRQ response time latency for deep package C-states. (Though others have the register, but do not use it) Lets print it out to give insight into the cases where it is used. IRTL begain in SNB, with PC3/PC6/PC7, and HSW added PC8/PC9/PC10. Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:55 +08:00
if (debug && do_irtl_snb)
print_irtl();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
int fork_it(char **argv)
{
pid_t child_pid;
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 12:08:19 +08:00
int status;
tools/power turbostat: fix un-intended affinity of forked program Linux 3.4 included a modification to turbostat to lower cross-call overhead by using scheduler affinity: 15aaa34654831e98dd76f7738b6c7f5d05a66430 (tools turbostat: reduce measurement overhead due to IPIs) In the use-case where turbostat forks a child program, that change had the un-intended side-effect of binding the child to the last cpu in the system. This change removed the binding before forking the child. This is a back-port of a fix already included in turbostat v2. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 11:24:00 +08:00
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 12:08:19 +08:00
status = for_all_cpus(get_counters, EVEN_COUNTERS);
if (status)
exit(status);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/* clear affinity side-effect of get_counters() */
sched_setaffinity(0, cpu_present_setsize, cpu_present_set);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
gettimeofday(&tv_even, (struct timezone *)NULL);
child_pid = fork();
if (!child_pid) {
/* child */
execvp(argv[0], argv);
} else {
/* parent */
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (child_pid == -1)
err(1, "fork");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
signal(SIGINT, SIG_IGN);
signal(SIGQUIT, SIG_IGN);
turbostat: Clean up error handling; disambiguate error messages; use err and errx Most of turbostat's error handling consists of printing an error (often including an errno) and exiting. Since perror doesn't support a format string, those error messages are often ambiguous, such as just showing a file path, which doesn't uniquely identify which call failed. turbostat already uses _GNU_SOURCE, so switch to the err and errx functions from err.h, which take a format string. Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com>
2013-08-21 08:20:18 +08:00
if (waitpid(child_pid, &status, 0) == -1)
err(status, "waitpid");
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
/*
* n.b. fork_it() does not check for errors from for_all_cpus()
* because re-starting is problematic when forking
*/
for_all_cpus(get_counters, ODD_COUNTERS);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
gettimeofday(&tv_odd, (struct timezone *)NULL);
timersub(&tv_odd, &tv_even, &tv_delta);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
for_all_cpus_2(delta_cpu, ODD_COUNTERS, EVEN_COUNTERS);
compute_average(EVEN_COUNTERS);
format_all_counters(EVEN_COUNTERS);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "%.6f sec\n", tv_delta.tv_sec + tv_delta.tv_usec/1000000.0);
flush_output_stderr();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
tools/power turbostat: graceful fail on garbage input When invald MSR's are specified on the command line, turbostat should simply print an error and exit. Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-01 12:08:19 +08:00
return status;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
int get_and_dump_counters(void)
{
int status;
status = for_all_cpus(get_counters, ODD_COUNTERS);
if (status)
return status;
status = for_all_cpus(dump_counters, ODD_COUNTERS);
if (status)
return status;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
flush_output_stdout();
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
return status;
}
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
void print_version() {
tools/power turbostat: initial SKX support SKX has a lot in common with HSX Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-04-07 05:15:58 +08:00
fprintf(outf, "turbostat version 4.12 5 Apr 2016"
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
" - Len Brown <lenb@kernel.org>\n");
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
void cmdline(int argc, char **argv)
{
int opt;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
int option_index = 0;
static struct option long_options[] = {
{"Counter", required_argument, 0, 'C'},
{"counter", required_argument, 0, 'c'},
{"Dump", no_argument, 0, 'D'},
{"debug", no_argument, 0, 'd'},
{"interval", required_argument, 0, 'i'},
{"help", no_argument, 0, 'h'},
{"Joules", no_argument, 0, 'J'},
{"MSR", required_argument, 0, 'M'},
{"msr", required_argument, 0, 'm'},
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
{"out", required_argument, 0, 'o'},
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
{"Package", no_argument, 0, 'p'},
{"processor", no_argument, 0, 'p'},
{"Summary", no_argument, 0, 'S'},
{"TCC", required_argument, 0, 'T'},
{"version", no_argument, 0, 'v' },
{0, 0, 0, 0 }
};
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
progname = argv[0];
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
while ((opt = getopt_long_only(argc, argv, "+C:c:Ddhi:JM:m:o:PpST:v",
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
long_options, &option_index)) != -1) {
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
switch (opt) {
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'C':
sscanf(optarg, "%x", &extra_delta_offset64);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'c':
sscanf(optarg, "%x", &extra_delta_offset32);
tools/power: turbostat v2 - re-write for efficiency Measuring large profoundly-idle configurations requires turbostat to be more lightweight. Otherwise, the operation of turbostat itself can interfere with the measurements. This re-write makes turbostat topology aware. Hardware is accessed in "topology order". Redundant hardware accesses are deleted. Redundant output is deleted. Also, output is buffered and local RDTSC use replaces remote MSR access for TSC. From a feature point of view, the output looks different since redundant figures are absent. Also, there are now -c and -p options -- to restrict output to the 1st thread in each core, and the 1st thread in each package, respectively. This is helpful to reduce output on big systems, where more detail than the "-s" system summary is desired. Finally, periodic mode output is now on stdout, not stderr. Turbostat v2 is also slightly more robust in handling run-time CPU online/offline events, as it now checks the actual map of on-line cpus rather than just the total number of on-line cpus. Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'D':
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
dump_only++;
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'd':
debug++;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'h':
default:
help();
exit(1);
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
case 'i':
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 11:44:48 +08:00
{
double interval = strtod(optarg, NULL);
if (interval < 0.001) {
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
fprintf(outf, "interval %f seconds is too small\n",
tools/power turbostat: allow sub-sec intervals turbostat -i interval_sec will sample and display statistics every interval_sec. interval_sec used to be a whole number of seconds, but now we accept a decimal, as small as 0.001 sec (1 ms). Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-13 11:44:48 +08:00
interval);
exit(2);
}
interval_ts.tv_sec = interval;
interval_ts.tv_nsec = (interval - interval_ts.tv_sec) * 1000000000;
}
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'J':
rapl_joules++;
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'M':
sscanf(optarg, "%x", &extra_msr_offset64);
tools/power turbostat: add [-d MSR#][-D MSR#] options to print counter deltas # turbostat -d 0x34 is useful for printing the number of SMI's within an interval on Nehalem and newer processors. where # turbostat -m 0x34 will simply print out the total SMI count since reset. Suggested-by: Andi Kleen Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 13:25:08 +08:00
break;
tools/power turbostat: add [-m MSR#] option -m MSR# prints the specified MSR in 32-bit format -M MSR# prints the specified MSR in 64-bit format Signed-off-by: Len Brown <len.brown@intel.com>
2012-09-22 11:45:46 +08:00
case 'm':
sscanf(optarg, "%x", &extra_msr_offset32);
break;
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
case 'o':
outf = fopen_or_die(optarg, "w");
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'P':
show_pkg_only++;
break;
case 'p':
show_core_only++;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'S':
summary_only++;
tools/power turbostat: v3.0: monitor Watts and Temperature Show power in Watts and temperature in Celsius when hardware support is present. Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL (Run-Time-Average-Power-Limiting). Per the Intel SDM (Intel® 64 and IA-32 Architectures Software Developer Manual) RAPL provides hardware energy counters and power control MSRs (Model Specific Registers). RAPL MSRs are designed primarily as a method to implement power capping. However, they are useful for monitoring system power whether or not power capping is used. In addition, Turbostat now shows temperature from DTS (Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware, if present. As before, turbostat reads MSRs, and never writes MSRs. New columns are present in turbostat output: The Pkg_W column shows Watts for each package (socket) in the system. On multi-socket systems, the system summary on the 1st row shows the sum for all sockets together. The Cor_W column shows Watts due to processors cores. Note that Core_W is included in Pkg_W. The optional GFX_W column shows Watts due to the graphics "un-core". Note that GFX_W is included in Pkg_W. The optional RAM_W column on server processors shows Watts due to DRAM DIMMS. As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W. The optional PKG_% and RAM_% columns on server processors shows the % of time in the measurement interval that RAPL power limiting is in effect on the package and on DRAM. Note that the RAPL energy counters have some limitations. First, hardware updates the counters about once every milli-second. This is fine for typical turbostat measurement intervals > 1 sec. However, when turbostat is used to measure events that approach 1ms, the counters are less useful. Second, the 32-bit energy counters are subject to wrapping. For example, a counter incrementing 15 micro-Joule units on a 130 Watt TDP server processor could (in theory) roll over in about 9 minutes. Turbostat detects and handles up to 1 counter overflow per measurement interval. But when the measurement interval exceeds the guaranteed counter range, we can't detect if more than 1 overflow occured. So in this case turbostat indicates that the results are in question by replacing the fractional part of the Watts in the output with "**": Pkg_W Cor_W GFX_W 3** 0** 0** Third, the RAPL counters are energy (Joule) counters -- they sum up weighted events in the package to estimate energy consumed. They are not analong power (Watt) meters. In practice, they tend to under-count because they don't cover every possible use of energy in the package. The accuracy of the RAPL counters will vary between product generations, and between SKU's in the same product generation, and with temperature. turbostat's -v (verbose) option now displays more power and thermal configuration information -- as shown on the turbostat.8 manual page. For example, it now displays the Package and DRAM Thermal Design Power (TDP): cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.) cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.) Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
break;
case 'T':
tcc_activation_temp_override = atoi(optarg);
break;
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
case 'v':
print_version();
exit(0);
turbostat: Add option to report joules consumed per sample Add "-J" option to report energy consumed in joules per sample. This option also adds the sample time to the reported values. Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2013-12-17 02:23:41 +08:00
break;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
}
}
}
int main(int argc, char **argv)
{
tools/power turbostat: add --out option for saving output in a file By default... Turbostat --debug gconfiguration info goes to stderr. In FORK mode, turbostat statistics go to stderr. In PERIODIC mode, turbostat statistics go to stdout. These defaults do not change, but an option "--out file" will send all output above only to the specified file. Signed-off-by: Len Brown <len.brown@intel.com>
2016-02-14 12:36:17 +08:00
outf = stderr;
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
cmdline(argc, argv);
tools/power turbostat: update parameters, documentation Long format options added, though the short ones should still work. eg. the new "--Counter 0x10" is the same as the old "-C 0x10" Note this Incompatibility: Old: -v displayed verbose debug output New: -v and --version simpaly display version Additional parameters: -d and --debug display verbose debug output -h and --help display a help message Updated turbosat.8 man page accordingly. Signed-off-by: Len Brown <len.brown@intel.com>
2015-02-10 14:56:38 +08:00
if (debug)
print_version();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
turbostat_init();
tools/power turbostat: introduce -s to dump counters The new option allows just run turbostat and get dump of counter values. It's useful when we have something more than one program to test. Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
2014-01-23 23:13:15 +08:00
/* dump counters and exit */
if (dump_only)
return get_and_dump_counters();
tools: create power/x86/turbostat turbostat is a Linux tool to observe proper operation of Intel(R) Turbo Boost Technology. turbostat displays the actual processor frequency on x86 processors that include APERF and MPERF MSRs. Note that turbostat is of limited utility on Linux kernels 2.6.29 and older, as acpi_cpufreq cleared APERF/MPERF up through that release. On Intel Core i3/i5/i7 (Nehalem) and newer processors, turbostat also displays residency in idle power saving states, which are necessary for diagnosing any cpuidle issues that may have an effect on turbo-mode. See the turbostat.8 man page for example usage. Signed-off-by: Len Brown <len.brown@intel.com>
2010-10-23 11:53:03 +08:00
/*
* if any params left, it must be a command to fork
*/
if (argc - optind)
return fork_it(argv + optind);
else
turbostat_loop();
return 0;
}
Reference in New Issue Copy Permalink