linux/tools/perf/util/callchain.c

1733 lines
40 KiB
C
Raw Normal View History

License cleanup: add SPDX GPL-2.0 license identifier to files with no license Many source files in the tree are missing licensing information, which makes it harder for compliance tools to determine the correct license. By default all files without license information are under the default license of the kernel, which is GPL version 2. Update the files which contain no license information with the 'GPL-2.0' SPDX license identifier. The SPDX identifier is a legally binding shorthand, which can be used instead of the full boiler plate text. This patch is based on work done by Thomas Gleixner and Kate Stewart and Philippe Ombredanne. How this work was done: Patches were generated and checked against linux-4.14-rc6 for a subset of the use cases: - file had no licensing information it it. - file was a */uapi/* one with no licensing information in it, - file was a */uapi/* one with existing licensing information, Further patches will be generated in subsequent months to fix up cases where non-standard license headers were used, and references to license had to be inferred by heuristics based on keywords. The analysis to determine which SPDX License Identifier to be applied to a file was done in a spreadsheet of side by side results from of the output of two independent scanners (ScanCode & Windriver) producing SPDX tag:value files created by Philippe Ombredanne. Philippe prepared the base worksheet, and did an initial spot review of a few 1000 files. The 4.13 kernel was the starting point of the analysis with 60,537 files assessed. Kate Stewart did a file by file comparison of the scanner results in the spreadsheet to determine which SPDX license identifier(s) to be applied to the file. She confirmed any determination that was not immediately clear with lawyers working with the Linux Foundation. Criteria used to select files for SPDX license identifier tagging was: - Files considered eligible had to be source code files. - Make and config files were included as candidates if they contained >5 lines of source - File already had some variant of a license header in it (even if <5 lines). All documentation files were explicitly excluded. The following heuristics were used to determine which SPDX license identifiers to apply. - when both scanners couldn't find any license traces, file was considered to have no license information in it, and the top level COPYING file license applied. For non */uapi/* files that summary was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 11139 and resulted in the first patch in this series. If that file was a */uapi/* path one, it was "GPL-2.0 WITH Linux-syscall-note" otherwise it was "GPL-2.0". Results of that was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 WITH Linux-syscall-note 930 and resulted in the second patch in this series. - if a file had some form of licensing information in it, and was one of the */uapi/* ones, it was denoted with the Linux-syscall-note if any GPL family license was found in the file or had no licensing in it (per prior point). Results summary: SPDX license identifier # files ---------------------------------------------------|------ GPL-2.0 WITH Linux-syscall-note 270 GPL-2.0+ WITH Linux-syscall-note 169 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) 21 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) 17 LGPL-2.1+ WITH Linux-syscall-note 15 GPL-1.0+ WITH Linux-syscall-note 14 ((GPL-2.0+ WITH Linux-syscall-note) OR BSD-3-Clause) 5 LGPL-2.0+ WITH Linux-syscall-note 4 LGPL-2.1 WITH Linux-syscall-note 3 ((GPL-2.0 WITH Linux-syscall-note) OR MIT) 3 ((GPL-2.0 WITH Linux-syscall-note) AND MIT) 1 and that resulted in the third patch in this series. - when the two scanners agreed on the detected license(s), that became the concluded license(s). - when there was disagreement between the two scanners (one detected a license but the other didn't, or they both detected different licenses) a manual inspection of the file occurred. - In most cases a manual inspection of the information in the file resulted in a clear resolution of the license that should apply (and which scanner probably needed to revisit its heuristics). - When it was not immediately clear, the license identifier was confirmed with lawyers working with the Linux Foundation. - If there was any question as to the appropriate license identifier, the file was flagged for further research and to be revisited later in time. In total, over 70 hours of logged manual review was done on the spreadsheet to determine the SPDX license identifiers to apply to the source files by Kate, Philippe, Thomas and, in some cases, confirmation by lawyers working with the Linux Foundation. Kate also obtained a third independent scan of the 4.13 code base from FOSSology, and compared selected files where the other two scanners disagreed against that SPDX file, to see if there was new insights. The Windriver scanner is based on an older version of FOSSology in part, so they are related. Thomas did random spot checks in about 500 files from the spreadsheets for the uapi headers and agreed with SPDX license identifier in the files he inspected. For the non-uapi files Thomas did random spot checks in about 15000 files. In initial set of patches against 4.14-rc6, 3 files were found to have copy/paste license identifier errors, and have been fixed to reflect the correct identifier. Additionally Philippe spent 10 hours this week doing a detailed manual inspection and review of the 12,461 patched files from the initial patch version early this week with: - a full scancode scan run, collecting the matched texts, detected license ids and scores - reviewing anything where there was a license detected (about 500+ files) to ensure that the applied SPDX license was correct - reviewing anything where there was no detection but the patch license was not GPL-2.0 WITH Linux-syscall-note to ensure that the applied SPDX license was correct This produced a worksheet with 20 files needing minor correction. This worksheet was then exported into 3 different .csv files for the different types of files to be modified. These .csv files were then reviewed by Greg. Thomas wrote a script to parse the csv files and add the proper SPDX tag to the file, in the format that the file expected. This script was further refined by Greg based on the output to detect more types of files automatically and to distinguish between header and source .c files (which need different comment types.) Finally Greg ran the script using the .csv files to generate the patches. Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Philippe Ombredanne <pombredanne@nexb.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-11-01 22:07:57 +08:00
// SPDX-License-Identifier: GPL-2.0
/*
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
* Copyright (C) 2009-2011, Frederic Weisbecker <fweisbec@gmail.com>
*
* Handle the callchains from the stream in an ad-hoc radix tree and then
* sort them in an rbtree.
*
* Using a radix for code path provides a fast retrieval and factorizes
* memory use. Also that lets us use the paths in a hierarchical graph view.
*
*/
#include <inttypes.h>
#include <stdlib.h>
#include <stdio.h>
#include <stdbool.h>
#include <errno.h>
#include <math.h>
#include <linux/string.h>
#include <linux/zalloc.h>
#include "asm/bug.h"
#include "debug.h"
#include "dso.h"
#include "event.h"
#include "hist.h"
#include "sort.h"
#include "machine.h"
#include "map.h"
#include "callchain.h"
#include "branch.h"
#include "symbol.h"
#include "util.h"
#include "../perf.h"
#define CALLCHAIN_PARAM_DEFAULT \
.mode = CHAIN_GRAPH_ABS, \
.min_percent = 0.5, \
.order = ORDER_CALLEE, \
.key = CCKEY_FUNCTION, \
.value = CCVAL_PERCENT, \
struct callchain_param callchain_param = {
CALLCHAIN_PARAM_DEFAULT
};
perf unwind: Do not look just at the global callchain_param.record_mode When setting up DWARF callchains on specific events, without using 'record' or 'trace' --call-graph, but instead doing it like: perf trace -e cycles/call-graph=dwarf/ The unwind__prepare_access() call in thread__insert_map() when we process PERF_RECORD_MMAP(2) metadata events were not being performed, precluding us from using per-event DWARF callchains, handling them just when we asked for all events to be DWARF, using "--call-graph dwarf". We do it in the PERF_RECORD_MMAP because we have to look at one of the executable maps to figure out the executable type (64-bit, 32-bit) of the DSO laid out in that mmap. Also to look at the architecture where the perf.data file was recorded. All this probably should be deferred to when we process a sample for some thread that has callchains, so that we do this processing only for the threads with samples, not for all of them. For now, fix using DWARF on specific events. Before: # perf trace --no-syscalls -e probe_libc:inet_pton/call-graph=dwarf/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.048 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.048/0.048/0.048/0.000 ms 0.000 probe_libc:inet_pton:(7fe9597bb350)) Problem processing probe_libc:inet_pton callchain, skipping... # After: # perf trace --no-syscalls -e probe_libc:inet_pton/call-graph=dwarf/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.060 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.060/0.060/0.060/0.000 ms 0.000 probe_libc:inet_pton:(7fd4aa930350)) __inet_pton (inlined) gaih_inet.constprop.7 (/usr/lib64/libc-2.26.so) __GI_getaddrinfo (inlined) [0xffffaa804e51af3f] (/usr/bin/ping) __libc_start_main (/usr/lib64/libc-2.26.so) [0xffffaa804e51b379] (/usr/bin/ping) # # perf trace --call-graph=dwarf --no-syscalls -e probe_libc:inet_pton/call-graph=dwarf/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.057 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.057/0.057/0.057/0.000 ms 0.000 probe_libc:inet_pton:(7f9363b9e350)) __inet_pton (inlined) gaih_inet.constprop.7 (/usr/lib64/libc-2.26.so) __GI_getaddrinfo (inlined) [0xffffa9e8a14e0f3f] (/usr/bin/ping) __libc_start_main (/usr/lib64/libc-2.26.so) [0xffffa9e8a14e1379] (/usr/bin/ping) # # perf trace --call-graph=fp --no-syscalls -e probe_libc:inet_pton/call-graph=dwarf/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.077 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.077/0.077/0.077/0.000 ms 0.000 probe_libc:inet_pton:(7f4947e1c350)) __inet_pton (inlined) gaih_inet.constprop.7 (/usr/lib64/libc-2.26.so) __GI_getaddrinfo (inlined) [0xffffaa716d88ef3f] (/usr/bin/ping) __libc_start_main (/usr/lib64/libc-2.26.so) [0xffffaa716d88f379] (/usr/bin/ping) # # perf trace --no-syscalls -e probe_libc:inet_pton/call-graph=fp/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.078 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.078/0.078/0.078/0.000 ms 0.000 probe_libc:inet_pton:(7fa157696350)) __GI___inet_pton (/usr/lib64/libc-2.26.so) getaddrinfo (/usr/lib64/libc-2.26.so) [0xffffa9ba39c74f40] (/usr/bin/ping) # Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Hendrick Brueckner <brueckner@linux.vnet.ibm.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Thomas Richter <tmricht@linux.vnet.ibm.com> Cc: Wang Nan <wangnan0@huawei.com> Link: https://lkml.kernel.org/r/20180116182650.GE16107@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2018-01-16 03:48:46 +08:00
/*
* Are there any events usind DWARF callchains?
*
* I.e.
*
* -e cycles/call-graph=dwarf/
*/
bool dwarf_callchain_users;
struct callchain_param callchain_param_default = {
CALLCHAIN_PARAM_DEFAULT
};
__thread struct callchain_cursor callchain_cursor;
int parse_callchain_record_opt(const char *arg, struct callchain_param *param)
{
return parse_callchain_record(arg, param);
}
static int parse_callchain_mode(const char *value)
{
if (!strncmp(value, "graph", strlen(value))) {
callchain_param.mode = CHAIN_GRAPH_ABS;
return 0;
}
if (!strncmp(value, "flat", strlen(value))) {
callchain_param.mode = CHAIN_FLAT;
return 0;
}
if (!strncmp(value, "fractal", strlen(value))) {
callchain_param.mode = CHAIN_GRAPH_REL;
return 0;
}
perf report: Support folded callchain mode on --stdio Add new call chain option (-g) 'folded' to print callchains in a line. The callchains are separated by semicolons, and preceded by (absolute) percent values and a space. For example, the following 20 lines can be printed in 3 lines with the folded output mode: $ perf report -g flat --no-children | grep -v ^# | head -20 60.48% swapper [kernel.vmlinux] [k] intel_idle 54.60% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary 5.88% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry rest_init start_kernel x86_64_start_reservations x86_64_start_kernel $ perf report -g folded --no-children | grep -v ^# | head -3 60.48% swapper [kernel.vmlinux] [k] intel_idle 54.60% intel_idle;cpuidle_enter_state;cpuidle_enter;call_cpuidle;cpu_startup_entry;start_secondary 5.88% intel_idle;cpuidle_enter_state;cpuidle_enter;call_cpuidle;cpu_startup_entry;rest_init;start_kernel;x86_64_start_reservations;x86_64_start_kernel This mode is supported only for --stdio now and intended to be used by some scripts like in FlameGraphs[1]. Support for other UI might be added later. [1] http://www.brendangregg.com/FlameGraphs/cpuflamegraphs.html Requested-and-Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:37 +08:00
if (!strncmp(value, "folded", strlen(value))) {
callchain_param.mode = CHAIN_FOLDED;
return 0;
}
return -1;
}
static int parse_callchain_order(const char *value)
{
if (!strncmp(value, "caller", strlen(value))) {
callchain_param.order = ORDER_CALLER;
callchain_param.order_set = true;
return 0;
}
if (!strncmp(value, "callee", strlen(value))) {
callchain_param.order = ORDER_CALLEE;
callchain_param.order_set = true;
return 0;
}
return -1;
}
static int parse_callchain_sort_key(const char *value)
{
if (!strncmp(value, "function", strlen(value))) {
callchain_param.key = CCKEY_FUNCTION;
return 0;
}
if (!strncmp(value, "address", strlen(value))) {
callchain_param.key = CCKEY_ADDRESS;
return 0;
}
perf report: Enable sorting by srcline as key Often it is interesting to know how costly a given source line is in total. Previously, one had to build these sums manually based on all addresses that pointed to the same source line. This patch introduces srcline as a sort key, which will do the aggregation for us. Paired with the recent addition of showing inline frames, this makes perf report much more useful for many C++ work loads. The following shows the new feature in action. First, let's show the status quo output when we sort by address. The result contains many hist entries that generate the same output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g address # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--60.31%--hypot +20 | | | | | |--8.52%--__hypot_finite +273 | | | | | |--7.32%--__hypot_finite +411 ... --35.34%--_start +4194346 __libc_start_main +241 | |--6.65%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--2.70%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--1.69%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ With this patch and `-g srcline` we instead get the following output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g srcline # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--64.02%--hypot | | | | | --59.81%--__hypot_finite | | | --0.53%--cabs | --35.34%--_start __libc_start_main | |--12.48%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20170318214928.9047-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-03-19 05:49:28 +08:00
if (!strncmp(value, "srcline", strlen(value))) {
callchain_param.key = CCKEY_SRCLINE;
return 0;
}
perf callchain: Support handling complete branch stacks as histograms Currently branch stacks can be only shown as edge histograms for individual branches. I never found this display particularly useful. This implements an alternative mode that creates histograms over complete branch traces, instead of individual branches, similar to how normal callgraphs are handled. This is done by putting it in front of the normal callgraph and then using the normal callgraph histogram infrastructure to unify them. This way in complex functions we can understand the control flow that lead to a particular sample, and may even see some control flow in the caller for short functions. Example (simplified, of course for such simple code this is usually not needed), please run this after the whole patchkit is in, as at this point in the patch order there is no --branch-history, that will be added in a patch after this one: tcall.c: volatile a = 10000, b = 100000, c; __attribute__((noinline)) f2() { c = a / b; } __attribute__((noinline)) f1() { f2(); f2(); } main() { int i; for (i = 0; i < 1000000; i++) f1(); } % perf record -b -g ./tsrc/tcall [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.044 MB perf.data (~1923 samples) ] % perf report --no-children --branch-history ... 54.91% tcall.c:6 [.] f2 tcall | |--65.53%-- f2 tcall.c:5 | | | |--70.83%-- f1 tcall.c:11 | | f1 tcall.c:10 | | main tcall.c:18 | | main tcall.c:18 | | main tcall.c:17 | | main tcall.c:17 | | f1 tcall.c:13 | | f1 tcall.c:13 | | f2 tcall.c:7 | | f2 tcall.c:5 | | f1 tcall.c:12 | | f1 tcall.c:12 | | f2 tcall.c:7 | | f2 tcall.c:5 | | f1 tcall.c:11 | | | --29.17%-- f1 tcall.c:12 | f1 tcall.c:12 | f2 tcall.c:7 | f2 tcall.c:5 | f1 tcall.c:11 | f1 tcall.c:10 | main tcall.c:18 | main tcall.c:18 | main tcall.c:17 | main tcall.c:17 | f1 tcall.c:13 | f1 tcall.c:13 | f2 tcall.c:7 | f2 tcall.c:5 | f1 tcall.c:12 The default output is unchanged. This is only implemented in perf report, no change to record or anywhere else. This adds the basic code to report: - add a new "branch" option to the -g option parser to enable this mode - when the flag is set include the LBR into the callstack in machine.c. The rest of the history code is unchanged and doesn't know the difference between LBR entry and normal call entry. - detect overlaps with the callchain - remove small loop duplicates in the LBR Current limitations: - The LBR flags (mispredict etc.) are not shown in the history and LBR entries have no special marker. - It would be nice if annotate marked the LBR entries somehow (e.g. with arrows) v2: Various fixes. v3: Merge further patches into this one. Fix white space. v4: Improve manpage. Address review feedback. v5: Rename functions. Better error message without -g. Fix crash without -b. v6: Rebase v7: Rebase. Use NO_ENTRY in memset. v8: Port to latest tip. Move add_callchain_ip to separate patch. Skip initial entries in callchain. Minor cleanups. Signed-off-by: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1415844328-4884-3-git-send-email-andi@firstfloor.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2014-11-13 10:05:20 +08:00
if (!strncmp(value, "branch", strlen(value))) {
callchain_param.branch_callstack = 1;
return 0;
}
return -1;
}
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
static int parse_callchain_value(const char *value)
{
if (!strncmp(value, "percent", strlen(value))) {
callchain_param.value = CCVAL_PERCENT;
return 0;
}
if (!strncmp(value, "period", strlen(value))) {
callchain_param.value = CCVAL_PERIOD;
return 0;
}
if (!strncmp(value, "count", strlen(value))) {
callchain_param.value = CCVAL_COUNT;
return 0;
}
return -1;
}
static int get_stack_size(const char *str, unsigned long *_size)
{
char *endptr;
unsigned long size;
unsigned long max_size = round_down(USHRT_MAX, sizeof(u64));
size = strtoul(str, &endptr, 0);
do {
if (*endptr)
break;
size = round_up(size, sizeof(u64));
if (!size || size > max_size)
break;
*_size = size;
return 0;
} while (0);
pr_err("callchain: Incorrect stack dump size (max %ld): %s\n",
max_size, str);
return -1;
}
static int
__parse_callchain_report_opt(const char *arg, bool allow_record_opt)
{
char *tok;
char *endptr, *saveptr = NULL;
bool minpcnt_set = false;
bool record_opt_set = false;
bool try_stack_size = false;
callchain_param.enabled = true;
symbol_conf.use_callchain = true;
if (!arg)
return 0;
while ((tok = strtok_r((char *)arg, ",", &saveptr)) != NULL) {
if (!strncmp(tok, "none", strlen(tok))) {
callchain_param.mode = CHAIN_NONE;
callchain_param.enabled = false;
symbol_conf.use_callchain = false;
return 0;
}
if (!parse_callchain_mode(tok) ||
!parse_callchain_order(tok) ||
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
!parse_callchain_sort_key(tok) ||
!parse_callchain_value(tok)) {
/* parsing ok - move on to the next */
try_stack_size = false;
goto next;
} else if (allow_record_opt && !record_opt_set) {
if (parse_callchain_record(tok, &callchain_param))
goto try_numbers;
/* assume that number followed by 'dwarf' is stack size */
if (callchain_param.record_mode == CALLCHAIN_DWARF)
try_stack_size = true;
record_opt_set = true;
goto next;
}
try_numbers:
if (try_stack_size) {
unsigned long size = 0;
if (get_stack_size(tok, &size) < 0)
return -1;
callchain_param.dump_size = size;
try_stack_size = false;
} else if (!minpcnt_set) {
/* try to get the min percent */
callchain_param.min_percent = strtod(tok, &endptr);
if (tok == endptr)
return -1;
minpcnt_set = true;
} else {
/* try print limit at last */
callchain_param.print_limit = strtoul(tok, &endptr, 0);
if (tok == endptr)
return -1;
}
next:
arg = NULL;
}
if (callchain_register_param(&callchain_param) < 0) {
pr_err("Can't register callchain params\n");
return -1;
}
return 0;
}
int parse_callchain_report_opt(const char *arg)
{
return __parse_callchain_report_opt(arg, false);
}
int parse_callchain_top_opt(const char *arg)
{
return __parse_callchain_report_opt(arg, true);
}
int parse_callchain_record(const char *arg, struct callchain_param *param)
{
char *tok, *name, *saveptr = NULL;
char *buf;
int ret = -1;
/* We need buffer that we know we can write to. */
buf = malloc(strlen(arg) + 1);
if (!buf)
return -ENOMEM;
strcpy(buf, arg);
tok = strtok_r((char *)buf, ",", &saveptr);
name = tok ? : (char *)buf;
do {
/* Framepointer style */
if (!strncmp(name, "fp", sizeof("fp"))) {
ret = 0;
param->record_mode = CALLCHAIN_FP;
tok = strtok_r(NULL, ",", &saveptr);
if (tok) {
unsigned long size;
size = strtoul(tok, &name, 0);
if (size < (unsigned) sysctl__max_stack())
param->max_stack = size;
}
break;
/* Dwarf style */
} else if (!strncmp(name, "dwarf", sizeof("dwarf"))) {
const unsigned long default_stack_dump_size = 8192;
ret = 0;
param->record_mode = CALLCHAIN_DWARF;
param->dump_size = default_stack_dump_size;
perf unwind: Do not look just at the global callchain_param.record_mode When setting up DWARF callchains on specific events, without using 'record' or 'trace' --call-graph, but instead doing it like: perf trace -e cycles/call-graph=dwarf/ The unwind__prepare_access() call in thread__insert_map() when we process PERF_RECORD_MMAP(2) metadata events were not being performed, precluding us from using per-event DWARF callchains, handling them just when we asked for all events to be DWARF, using "--call-graph dwarf". We do it in the PERF_RECORD_MMAP because we have to look at one of the executable maps to figure out the executable type (64-bit, 32-bit) of the DSO laid out in that mmap. Also to look at the architecture where the perf.data file was recorded. All this probably should be deferred to when we process a sample for some thread that has callchains, so that we do this processing only for the threads with samples, not for all of them. For now, fix using DWARF on specific events. Before: # perf trace --no-syscalls -e probe_libc:inet_pton/call-graph=dwarf/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.048 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.048/0.048/0.048/0.000 ms 0.000 probe_libc:inet_pton:(7fe9597bb350)) Problem processing probe_libc:inet_pton callchain, skipping... # After: # perf trace --no-syscalls -e probe_libc:inet_pton/call-graph=dwarf/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.060 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.060/0.060/0.060/0.000 ms 0.000 probe_libc:inet_pton:(7fd4aa930350)) __inet_pton (inlined) gaih_inet.constprop.7 (/usr/lib64/libc-2.26.so) __GI_getaddrinfo (inlined) [0xffffaa804e51af3f] (/usr/bin/ping) __libc_start_main (/usr/lib64/libc-2.26.so) [0xffffaa804e51b379] (/usr/bin/ping) # # perf trace --call-graph=dwarf --no-syscalls -e probe_libc:inet_pton/call-graph=dwarf/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.057 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.057/0.057/0.057/0.000 ms 0.000 probe_libc:inet_pton:(7f9363b9e350)) __inet_pton (inlined) gaih_inet.constprop.7 (/usr/lib64/libc-2.26.so) __GI_getaddrinfo (inlined) [0xffffa9e8a14e0f3f] (/usr/bin/ping) __libc_start_main (/usr/lib64/libc-2.26.so) [0xffffa9e8a14e1379] (/usr/bin/ping) # # perf trace --call-graph=fp --no-syscalls -e probe_libc:inet_pton/call-graph=dwarf/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.077 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.077/0.077/0.077/0.000 ms 0.000 probe_libc:inet_pton:(7f4947e1c350)) __inet_pton (inlined) gaih_inet.constprop.7 (/usr/lib64/libc-2.26.so) __GI_getaddrinfo (inlined) [0xffffaa716d88ef3f] (/usr/bin/ping) __libc_start_main (/usr/lib64/libc-2.26.so) [0xffffaa716d88f379] (/usr/bin/ping) # # perf trace --no-syscalls -e probe_libc:inet_pton/call-graph=fp/ ping -6 -c 1 ::1 PING ::1(::1) 56 data bytes 64 bytes from ::1: icmp_seq=1 ttl=64 time=0.078 ms --- ::1 ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 0.078/0.078/0.078/0.000 ms 0.000 probe_libc:inet_pton:(7fa157696350)) __GI___inet_pton (/usr/lib64/libc-2.26.so) getaddrinfo (/usr/lib64/libc-2.26.so) [0xffffa9ba39c74f40] (/usr/bin/ping) # Acked-by: Namhyung Kim <namhyung@kernel.org> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Hendrick Brueckner <brueckner@linux.vnet.ibm.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Thomas Richter <tmricht@linux.vnet.ibm.com> Cc: Wang Nan <wangnan0@huawei.com> Link: https://lkml.kernel.org/r/20180116182650.GE16107@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2018-01-16 03:48:46 +08:00
dwarf_callchain_users = true;
tok = strtok_r(NULL, ",", &saveptr);
if (tok) {
unsigned long size = 0;
ret = get_stack_size(tok, &size);
param->dump_size = size;
}
} else if (!strncmp(name, "lbr", sizeof("lbr"))) {
if (!strtok_r(NULL, ",", &saveptr)) {
param->record_mode = CALLCHAIN_LBR;
ret = 0;
} else
pr_err("callchain: No more arguments "
"needed for --call-graph lbr\n");
break;
} else {
pr_err("callchain: Unknown --call-graph option "
"value: %s\n", arg);
break;
}
} while (0);
free(buf);
return ret;
}
int perf_callchain_config(const char *var, const char *value)
{
char *endptr;
if (!strstarts(var, "call-graph."))
return 0;
var += sizeof("call-graph.") - 1;
if (!strcmp(var, "record-mode"))
return parse_callchain_record_opt(value, &callchain_param);
if (!strcmp(var, "dump-size")) {
unsigned long size = 0;
int ret;
ret = get_stack_size(value, &size);
callchain_param.dump_size = size;
return ret;
}
perf report: Fix debug messages with --call-graph option With --call-graph option, perf report can display call chains using type, min percent threshold, optional print limit and order. And the default call-graph parameter is 'graph,0.5,caller,function,percent'. Before this patch, 'perf report --call-graph' shows incorrect debug messages as below: # perf report --call-graph Invalid callchain mode: 0.5 Invalid callchain order: 0.5 Invalid callchain sort key: 0.5 Invalid callchain config key: 0.5 Invalid callchain mode: caller Invalid callchain mode: function Invalid callchain order: function Invalid callchain mode: percent Invalid callchain order: percent Invalid callchain sort key: percent That is because in function __parse_callchain_report_opt(),each field of the call-graph parameter is passed to parse_callchain_{mode,order, sort_key,value} in turn until it meets the matching value. For example, the order field "caller" is passed to parse_callchain_mode() firstly and obviously it doesn't match any mode field. Therefore parse_callchain_mode() will shows the debug message "Invalid callchain mode: caller", which could confuse users. The patch fixes this issue by moving the warning out of the function parse_callchain_{mode,order,sort_key,value}. Signed-off-by: Mengting Zhang <zhangmengting@huawei.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Krister Johansen <kjlx@templeofstupid.com> Cc: Li Bin <huawei.libin@huawei.com> Cc: Milian Wolff <milian.wolff@kdab.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Wang Nan <wangnan0@huawei.com> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/1506154694-39691-1-git-send-email-zhangmengting@huawei.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-09-23 16:18:14 +08:00
if (!strcmp(var, "print-type")){
int ret;
ret = parse_callchain_mode(value);
if (ret == -1)
pr_err("Invalid callchain mode: %s\n", value);
return ret;
}
if (!strcmp(var, "order")){
int ret;
ret = parse_callchain_order(value);
if (ret == -1)
pr_err("Invalid callchain order: %s\n", value);
return ret;
}
if (!strcmp(var, "sort-key")){
int ret;
ret = parse_callchain_sort_key(value);
if (ret == -1)
pr_err("Invalid callchain sort key: %s\n", value);
return ret;
}
if (!strcmp(var, "threshold")) {
callchain_param.min_percent = strtod(value, &endptr);
perf tools: Propagate perf_config() errors Previously these were being ignored, sometimes silently. Stop doing that, emitting debug messages and handling the errors. Testing it: $ cat ~/.perfconfig cat: /home/acme/.perfconfig: No such file or directory $ perf stat -e cycles usleep 1 Performance counter stats for 'usleep 1': 938,996 cycles:u 0.003813731 seconds time elapsed $ perf top --stdio Error: You may not have permission to collect system-wide stats. Consider tweaking /proc/sys/kernel/perf_event_paranoid, <SNIP> [ perf record: Captured and wrote 0.019 MB perf.data (7 samples) ] [acme@jouet linux]$ perf report --stdio # To display the perf.data header info, please use --header/--header-only options. # Overhead Command Shared Object Symbol # ........ ....... ................. ......................... 71.77% usleep libc-2.24.so [.] _dl_addr 27.07% usleep ld-2.24.so [.] _dl_next_ld_env_entry 1.13% usleep [kernel.kallsyms] [k] page_fault $ $ touch ~/.perfconfig $ ls -la ~/.perfconfig -rw-rw-r--. 1 acme acme 0 Jan 27 12:14 /home/acme/.perfconfig $ $ perf stat -e instructions usleep 1 Performance counter stats for 'usleep 1': 244,610 instructions:u 0.000805383 seconds time elapsed $ [root@jouet ~]# chown acme.acme ~/.perfconfig [root@jouet ~]# perf stat -e cycles usleep 1 Warning: File /root/.perfconfig not owned by current user or root, ignoring it. Performance counter stats for 'usleep 1': 937,615 cycles 0.000836931 seconds time elapsed # Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/n/tip-j2rq96so6xdqlr8p8rd6a3jx@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-01-25 00:44:10 +08:00
if (value == endptr) {
pr_err("Invalid callchain threshold: %s\n", value);
return -1;
perf tools: Propagate perf_config() errors Previously these were being ignored, sometimes silently. Stop doing that, emitting debug messages and handling the errors. Testing it: $ cat ~/.perfconfig cat: /home/acme/.perfconfig: No such file or directory $ perf stat -e cycles usleep 1 Performance counter stats for 'usleep 1': 938,996 cycles:u 0.003813731 seconds time elapsed $ perf top --stdio Error: You may not have permission to collect system-wide stats. Consider tweaking /proc/sys/kernel/perf_event_paranoid, <SNIP> [ perf record: Captured and wrote 0.019 MB perf.data (7 samples) ] [acme@jouet linux]$ perf report --stdio # To display the perf.data header info, please use --header/--header-only options. # Overhead Command Shared Object Symbol # ........ ....... ................. ......................... 71.77% usleep libc-2.24.so [.] _dl_addr 27.07% usleep ld-2.24.so [.] _dl_next_ld_env_entry 1.13% usleep [kernel.kallsyms] [k] page_fault $ $ touch ~/.perfconfig $ ls -la ~/.perfconfig -rw-rw-r--. 1 acme acme 0 Jan 27 12:14 /home/acme/.perfconfig $ $ perf stat -e instructions usleep 1 Performance counter stats for 'usleep 1': 244,610 instructions:u 0.000805383 seconds time elapsed $ [root@jouet ~]# chown acme.acme ~/.perfconfig [root@jouet ~]# perf stat -e cycles usleep 1 Warning: File /root/.perfconfig not owned by current user or root, ignoring it. Performance counter stats for 'usleep 1': 937,615 cycles 0.000836931 seconds time elapsed # Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/n/tip-j2rq96so6xdqlr8p8rd6a3jx@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-01-25 00:44:10 +08:00
}
}
if (!strcmp(var, "print-limit")) {
callchain_param.print_limit = strtod(value, &endptr);
perf tools: Propagate perf_config() errors Previously these were being ignored, sometimes silently. Stop doing that, emitting debug messages and handling the errors. Testing it: $ cat ~/.perfconfig cat: /home/acme/.perfconfig: No such file or directory $ perf stat -e cycles usleep 1 Performance counter stats for 'usleep 1': 938,996 cycles:u 0.003813731 seconds time elapsed $ perf top --stdio Error: You may not have permission to collect system-wide stats. Consider tweaking /proc/sys/kernel/perf_event_paranoid, <SNIP> [ perf record: Captured and wrote 0.019 MB perf.data (7 samples) ] [acme@jouet linux]$ perf report --stdio # To display the perf.data header info, please use --header/--header-only options. # Overhead Command Shared Object Symbol # ........ ....... ................. ......................... 71.77% usleep libc-2.24.so [.] _dl_addr 27.07% usleep ld-2.24.so [.] _dl_next_ld_env_entry 1.13% usleep [kernel.kallsyms] [k] page_fault $ $ touch ~/.perfconfig $ ls -la ~/.perfconfig -rw-rw-r--. 1 acme acme 0 Jan 27 12:14 /home/acme/.perfconfig $ $ perf stat -e instructions usleep 1 Performance counter stats for 'usleep 1': 244,610 instructions:u 0.000805383 seconds time elapsed $ [root@jouet ~]# chown acme.acme ~/.perfconfig [root@jouet ~]# perf stat -e cycles usleep 1 Warning: File /root/.perfconfig not owned by current user or root, ignoring it. Performance counter stats for 'usleep 1': 937,615 cycles 0.000836931 seconds time elapsed # Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/n/tip-j2rq96so6xdqlr8p8rd6a3jx@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-01-25 00:44:10 +08:00
if (value == endptr) {
pr_err("Invalid callchain print limit: %s\n", value);
return -1;
perf tools: Propagate perf_config() errors Previously these were being ignored, sometimes silently. Stop doing that, emitting debug messages and handling the errors. Testing it: $ cat ~/.perfconfig cat: /home/acme/.perfconfig: No such file or directory $ perf stat -e cycles usleep 1 Performance counter stats for 'usleep 1': 938,996 cycles:u 0.003813731 seconds time elapsed $ perf top --stdio Error: You may not have permission to collect system-wide stats. Consider tweaking /proc/sys/kernel/perf_event_paranoid, <SNIP> [ perf record: Captured and wrote 0.019 MB perf.data (7 samples) ] [acme@jouet linux]$ perf report --stdio # To display the perf.data header info, please use --header/--header-only options. # Overhead Command Shared Object Symbol # ........ ....... ................. ......................... 71.77% usleep libc-2.24.so [.] _dl_addr 27.07% usleep ld-2.24.so [.] _dl_next_ld_env_entry 1.13% usleep [kernel.kallsyms] [k] page_fault $ $ touch ~/.perfconfig $ ls -la ~/.perfconfig -rw-rw-r--. 1 acme acme 0 Jan 27 12:14 /home/acme/.perfconfig $ $ perf stat -e instructions usleep 1 Performance counter stats for 'usleep 1': 244,610 instructions:u 0.000805383 seconds time elapsed $ [root@jouet ~]# chown acme.acme ~/.perfconfig [root@jouet ~]# perf stat -e cycles usleep 1 Warning: File /root/.perfconfig not owned by current user or root, ignoring it. Performance counter stats for 'usleep 1': 937,615 cycles 0.000836931 seconds time elapsed # Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/n/tip-j2rq96so6xdqlr8p8rd6a3jx@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-01-25 00:44:10 +08:00
}
}
return 0;
}
static void
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
rb_insert_callchain(struct rb_root *root, struct callchain_node *chain,
enum chain_mode mode)
{
struct rb_node **p = &root->rb_node;
struct rb_node *parent = NULL;
struct callchain_node *rnode;
u64 chain_cumul = callchain_cumul_hits(chain);
while (*p) {
u64 rnode_cumul;
parent = *p;
rnode = rb_entry(parent, struct callchain_node, rb_node);
rnode_cumul = callchain_cumul_hits(rnode);
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
switch (mode) {
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
case CHAIN_FLAT:
perf report: Support folded callchain mode on --stdio Add new call chain option (-g) 'folded' to print callchains in a line. The callchains are separated by semicolons, and preceded by (absolute) percent values and a space. For example, the following 20 lines can be printed in 3 lines with the folded output mode: $ perf report -g flat --no-children | grep -v ^# | head -20 60.48% swapper [kernel.vmlinux] [k] intel_idle 54.60% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary 5.88% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry rest_init start_kernel x86_64_start_reservations x86_64_start_kernel $ perf report -g folded --no-children | grep -v ^# | head -3 60.48% swapper [kernel.vmlinux] [k] intel_idle 54.60% intel_idle;cpuidle_enter_state;cpuidle_enter;call_cpuidle;cpu_startup_entry;start_secondary 5.88% intel_idle;cpuidle_enter_state;cpuidle_enter;call_cpuidle;cpu_startup_entry;rest_init;start_kernel;x86_64_start_reservations;x86_64_start_kernel This mode is supported only for --stdio now and intended to be used by some scripts like in FlameGraphs[1]. Support for other UI might be added later. [1] http://www.brendangregg.com/FlameGraphs/cpuflamegraphs.html Requested-and-Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:37 +08:00
case CHAIN_FOLDED:
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
if (rnode->hit < chain->hit)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
break;
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
case CHAIN_GRAPH_ABS: /* Falldown */
case CHAIN_GRAPH_REL:
if (rnode_cumul < chain_cumul)
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
break;
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
case CHAIN_NONE:
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
default:
break;
}
}
rb_link_node(&chain->rb_node, parent, p);
rb_insert_color(&chain->rb_node, root);
}
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
static void
__sort_chain_flat(struct rb_root *rb_root, struct callchain_node *node,
u64 min_hit)
{
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
struct rb_node *n;
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
struct callchain_node *child;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
n = rb_first(&node->rb_root_in);
while (n) {
child = rb_entry(n, struct callchain_node, rb_node_in);
n = rb_next(n);
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
__sort_chain_flat(rb_root, child, min_hit);
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
}
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
if (node->hit && node->hit >= min_hit)
rb_insert_callchain(rb_root, node, CHAIN_FLAT);
}
/*
* Once we get every callchains from the stream, we can now
* sort them by hit
*/
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
static void
sort_chain_flat(struct rb_root *rb_root, struct callchain_root *root,
perf tools: Use __maybe_used for unused variables perf defines both __used and __unused variables to use for marking unused variables. The variable __used is defined to __attribute__((__unused__)), which contradicts the kernel definition to __attribute__((__used__)) for new gcc versions. On Android, __used is also defined in system headers and this leads to warnings like: warning: '__used__' attribute ignored __unused is not defined in the kernel and is not a standard definition. If __unused is included everywhere instead of __used, this leads to conflicts with glibc headers, since glibc has a variables with this name in its headers. The best approach is to use __maybe_unused, the definition used in the kernel for __attribute__((unused)). In this way there is only one definition in perf sources (instead of 2 definitions that point to the same thing: __used and __unused) and it works on both Linux and Android. This patch simply replaces all instances of __used and __unused with __maybe_unused. Signed-off-by: Irina Tirdea <irina.tirdea@intel.com> Acked-by: Pekka Enberg <penberg@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1347315303-29906-7-git-send-email-irina.tirdea@intel.com [ committer note: fixed up conflict with a116e05 in builtin-sched.c ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-11 06:15:03 +08:00
u64 min_hit, struct callchain_param *param __maybe_unused)
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
{
*rb_root = RB_ROOT;
__sort_chain_flat(rb_root, &root->node, min_hit);
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
}
static void __sort_chain_graph_abs(struct callchain_node *node,
u64 min_hit)
{
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
struct rb_node *n;
struct callchain_node *child;
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
node->rb_root = RB_ROOT;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
n = rb_first(&node->rb_root_in);
while (n) {
child = rb_entry(n, struct callchain_node, rb_node_in);
n = rb_next(n);
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
__sort_chain_graph_abs(child, min_hit);
if (callchain_cumul_hits(child) >= min_hit)
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
rb_insert_callchain(&node->rb_root, child,
CHAIN_GRAPH_ABS);
}
}
static void
sort_chain_graph_abs(struct rb_root *rb_root, struct callchain_root *chain_root,
perf tools: Use __maybe_used for unused variables perf defines both __used and __unused variables to use for marking unused variables. The variable __used is defined to __attribute__((__unused__)), which contradicts the kernel definition to __attribute__((__used__)) for new gcc versions. On Android, __used is also defined in system headers and this leads to warnings like: warning: '__used__' attribute ignored __unused is not defined in the kernel and is not a standard definition. If __unused is included everywhere instead of __used, this leads to conflicts with glibc headers, since glibc has a variables with this name in its headers. The best approach is to use __maybe_unused, the definition used in the kernel for __attribute__((unused)). In this way there is only one definition in perf sources (instead of 2 definitions that point to the same thing: __used and __unused) and it works on both Linux and Android. This patch simply replaces all instances of __used and __unused with __maybe_unused. Signed-off-by: Irina Tirdea <irina.tirdea@intel.com> Acked-by: Pekka Enberg <penberg@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1347315303-29906-7-git-send-email-irina.tirdea@intel.com [ committer note: fixed up conflict with a116e05 in builtin-sched.c ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-11 06:15:03 +08:00
u64 min_hit, struct callchain_param *param __maybe_unused)
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
{
__sort_chain_graph_abs(&chain_root->node, min_hit);
rb_root->rb_node = chain_root->node.rb_root.rb_node;
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
}
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
static void __sort_chain_graph_rel(struct callchain_node *node,
double min_percent)
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
{
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
struct rb_node *n;
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
struct callchain_node *child;
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
u64 min_hit;
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
node->rb_root = RB_ROOT;
min_hit = ceil(node->children_hit * min_percent);
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
n = rb_first(&node->rb_root_in);
while (n) {
child = rb_entry(n, struct callchain_node, rb_node_in);
n = rb_next(n);
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
__sort_chain_graph_rel(child, min_percent);
if (callchain_cumul_hits(child) >= min_hit)
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
rb_insert_callchain(&node->rb_root, child,
CHAIN_GRAPH_REL);
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
}
}
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
static void
sort_chain_graph_rel(struct rb_root *rb_root, struct callchain_root *chain_root,
perf tools: Use __maybe_used for unused variables perf defines both __used and __unused variables to use for marking unused variables. The variable __used is defined to __attribute__((__unused__)), which contradicts the kernel definition to __attribute__((__used__)) for new gcc versions. On Android, __used is also defined in system headers and this leads to warnings like: warning: '__used__' attribute ignored __unused is not defined in the kernel and is not a standard definition. If __unused is included everywhere instead of __used, this leads to conflicts with glibc headers, since glibc has a variables with this name in its headers. The best approach is to use __maybe_unused, the definition used in the kernel for __attribute__((unused)). In this way there is only one definition in perf sources (instead of 2 definitions that point to the same thing: __used and __unused) and it works on both Linux and Android. This patch simply replaces all instances of __used and __unused with __maybe_unused. Signed-off-by: Irina Tirdea <irina.tirdea@intel.com> Acked-by: Pekka Enberg <penberg@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1347315303-29906-7-git-send-email-irina.tirdea@intel.com [ committer note: fixed up conflict with a116e05 in builtin-sched.c ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-11 06:15:03 +08:00
u64 min_hit __maybe_unused, struct callchain_param *param)
perf report: Add support for callchain graph output Currently, the printing of callchains is done in a single vertical level, this is the "flat" mode: 8.25% [k] copy_user_generic_string 4.19% copy_user_generic_string generic_file_aio_read do_sync_read vfs_read sys_pread64 system_call_fastpath pread64 This patch introduces a new "graph" mode which provides a hierarchical output of factorized paths recursively sorted: 8.25% [k] copy_user_generic_string | |--4.31%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--4.19%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --0.12%-- sys_read | system_call_fastpath | __read | |--3.24%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--3.14%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --0.10%-- sys_write [...] The command line has then changed. By providing the -c option, the callchain will output in the flat mode by default. But you can override it: perf report -c graph or perf report -c flat You can also pass the abreviated mode: perf report -c g or perf report -c gra will both make use of the graph mode. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-02 23:58:21 +08:00
{
__sort_chain_graph_rel(&chain_root->node, param->min_percent / 100.0);
rb_root->rb_node = chain_root->node.rb_root.rb_node;
}
int callchain_register_param(struct callchain_param *param)
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
{
switch (param->mode) {
case CHAIN_GRAPH_ABS:
param->sort = sort_chain_graph_abs;
break;
case CHAIN_GRAPH_REL:
param->sort = sort_chain_graph_rel;
break;
case CHAIN_FLAT:
perf report: Support folded callchain mode on --stdio Add new call chain option (-g) 'folded' to print callchains in a line. The callchains are separated by semicolons, and preceded by (absolute) percent values and a space. For example, the following 20 lines can be printed in 3 lines with the folded output mode: $ perf report -g flat --no-children | grep -v ^# | head -20 60.48% swapper [kernel.vmlinux] [k] intel_idle 54.60% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary 5.88% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry rest_init start_kernel x86_64_start_reservations x86_64_start_kernel $ perf report -g folded --no-children | grep -v ^# | head -3 60.48% swapper [kernel.vmlinux] [k] intel_idle 54.60% intel_idle;cpuidle_enter_state;cpuidle_enter;call_cpuidle;cpu_startup_entry;start_secondary 5.88% intel_idle;cpuidle_enter_state;cpuidle_enter;call_cpuidle;cpu_startup_entry;rest_init;start_kernel;x86_64_start_reservations;x86_64_start_kernel This mode is supported only for --stdio now and intended to be used by some scripts like in FlameGraphs[1]. Support for other UI might be added later. [1] http://www.brendangregg.com/FlameGraphs/cpuflamegraphs.html Requested-and-Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:37 +08:00
case CHAIN_FOLDED:
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
param->sort = sort_chain_flat;
break;
perf: Enable more compiler warnings Related to a shadowed variable bug fix Valdis Kletnieks noticed that perf does not get built with -Wshadow, which could have helped us avoid the bug. So enable -Wshadow and also enable the following warnings on perf builds, in addition to the already enabled -Wall -Wextra -std=gnu99 warnings: -Wcast-align -Wformat=2 -Wshadow -Winit-self -Wpacked -Wredundant-decls -Wstack-protector -Wstrict-aliasing=3 -Wswitch-default -Wswitch-enum -Wno-system-headers -Wundef -Wvolatile-register-var -Wwrite-strings -Wbad-function-cast -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wstrict-prototypes -Wdeclaration-after-statement And change/fix the perf code to build cleanly under GCC 4.3.2. The list of warnings enablement is rather arbitrary: it's based on my (quick) reading of the GCC manpages and trying them on perf. I categorized the warnings based on individually enabling them and looking whether they trigger something in the perf build. If i liked those warnings (i.e. if they trigger for something that arguably could be improved) i enabled the warning. If the warnings seemed to come from language laywers spamming the build with tons of nuisance warnings i generally kept them off. Most of the sign conversion related warnings were in this category. (A second patch enabling some of the sign warnings might be welcome - sign bugs can be nasty.) I also kept warnings that seem to make sense from their manpage description and which produced no actual warnings on our code base. These warnings might still be turned off if they end up being a nuisance. I also left out a few warnings that are not supported in older compilers. [ Note that these changes might break the build on older compilers i did not test, or on non-x86 architectures that produce different warnings, so more testing would be welcome. ] Reported-by: Valdis.Kletnieks@vt.edu Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-15 18:26:57 +08:00
case CHAIN_NONE:
perf report: Add "Fractal" mode output - support callchains with relative overhead rate The current callchain displays the overhead rates as absolute: relative to the total overhead. This patch provides relative overhead percentage, in which each branch of the callchain tree is a independant instrumentated object. This provides a 'fractal' view of the call-chain profile: each sub-graph looks like a profile in itself - relative to its parent. You can produce such output by using the "fractal" mode that you can abbreviate via f, fr, fra, frac, etc... ./perf report -s sym -c fractal Example: 8.46% [k] copy_user_generic_string | |--52.01%-- generic_file_aio_read | do_sync_read | vfs_read | | | |--97.20%-- sys_pread64 | | system_call_fastpath | | pread64 | | | --2.81%-- sys_read | system_call_fastpath | __read | |--39.85%-- generic_file_buffered_write | __generic_file_aio_write_nolock | generic_file_aio_write | do_sync_write | reiserfs_file_write | vfs_write | | | |--97.05%-- sys_pwrite64 | | system_call_fastpath | | __pwrite64 | | | --2.95%-- sys_write | system_call_fastpath | __write_nocancel [...] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> Cc: Jens Axboe <jens.axboe@oracle.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-05 13:39:21 +08:00
default:
return -1;
}
return 0;
}
/*
* Create a child for a parent. If inherit_children, then the new child
* will become the new parent of it's parent children
*/
static struct callchain_node *
create_child(struct callchain_node *parent, bool inherit_children)
{
struct callchain_node *new;
new = zalloc(sizeof(*new));
if (!new) {
perror("not enough memory to create child for code path tree");
return NULL;
}
new->parent = parent;
INIT_LIST_HEAD(&new->val);
perf hists browser: Support flat callchains The flat callchain mode is to print all chains in a single, simple hierarchy so make it easy to see. Currently perf report --tui doesn't show flat callchains properly. With flat callchains, only leaf nodes are added to the final rbtree so it should show entries in parent nodes. To do that, add parent_val list to struct callchain_node and show them along with the (normal) val list. For example, consider following callchains with '-g graph'. $ perf report -g graph - 39.93% swapper [kernel.vmlinux] [k] intel_idle intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle - cpu_startup_entry 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel Before: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel After: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle - 28.63% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary - 11.30% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_kernel x86_64_start_reservations x86_64_start_kernel Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-8-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:43 +08:00
INIT_LIST_HEAD(&new->parent_val);
if (inherit_children) {
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
struct rb_node *n;
struct callchain_node *child;
new->rb_root_in = parent->rb_root_in;
parent->rb_root_in = RB_ROOT;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
n = rb_first(&new->rb_root_in);
while (n) {
child = rb_entry(n, struct callchain_node, rb_node_in);
child->parent = new;
n = rb_next(n);
}
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
/* make it the first child */
rb_link_node(&new->rb_node_in, NULL, &parent->rb_root_in.rb_node);
rb_insert_color(&new->rb_node_in, &parent->rb_root_in);
}
return new;
}
perf: Fix orphan callchain branches Callchains have markers inside their capture to tell we enter a context (kernel, user, ...). Those are not displayed in the callchains but they are incidentally an active part of the radix tree where callchains are stored, just like any other address. If we have the two following callchains: addr1 -> addr2 -> user context -> addr3 addr1 -> addr2 -> user context -> addr4 addr1 -> addr2 -> addr 5 This is pretty common if addr1 and addr2 are part of an interrupt path, addr3 and addr4 are user addresses and addr5 is a kernel non interrupt path. This will be stored as follows in the tree: addr1 addr2 / \ / addr5 user context / \ addr3 addr4 But we ignore the context markers in the report, hence the addr3 and addr4 will appear as orphan branches: |--28.30%-- hrtimer_interrupt | smp_apic_timer_interrupt | apic_timer_interrupt | | <------------- here, no parent! | | | | | |--11.11%-- 0x7fae7bccb875 | | | | | |--11.11%-- 0xffffffffff60013b | | | | | |--11.11%-- __pthread_mutex_lock_internal | | | | | |--11.11%-- __errno_location Fix this by removing the context markers when we process the callchains to the tree. Reported-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1269274173-20328-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-23 00:09:33 +08:00
/*
* Fill the node with callchain values
*/
static int
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
fill_node(struct callchain_node *node, struct callchain_cursor *cursor)
{
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
struct callchain_cursor_node *cursor_node;
node->val_nr = cursor->nr - cursor->pos;
if (!node->val_nr)
pr_warning("Warning: empty node in callchain tree\n");
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
cursor_node = callchain_cursor_current(cursor);
while (cursor_node) {
struct callchain_list *call;
call = zalloc(sizeof(*call));
if (!call) {
perror("not enough memory for the code path tree");
return -1;
}
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
call->ip = cursor_node->ip;
call->ms = cursor_node->ms;
map__get(call->ms.map);
call->srcline = cursor_node->srcline;
if (cursor_node->branch) {
call->branch_count = 1;
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
if (cursor_node->branch_from) {
/*
* branch_from is set with value somewhere else
* to imply it's "to" of a branch.
*/
call->brtype_stat.branch_to = true;
if (cursor_node->branch_flags.predicted)
call->predicted_count = 1;
if (cursor_node->branch_flags.abort)
call->abort_count = 1;
branch_type_count(&call->brtype_stat,
&cursor_node->branch_flags,
cursor_node->branch_from,
cursor_node->ip);
} else {
/*
* It's "from" of a branch
*/
call->brtype_stat.branch_to = false;
call->cycles_count =
cursor_node->branch_flags.cycles;
call->iter_count = cursor_node->nr_loop_iter;
call->iter_cycles = cursor_node->iter_cycles;
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
}
}
list_add_tail(&call->list, &node->val);
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
callchain_cursor_advance(cursor);
cursor_node = callchain_cursor_current(cursor);
}
return 0;
}
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
static struct callchain_node *
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
add_child(struct callchain_node *parent,
struct callchain_cursor *cursor,
u64 period)
{
struct callchain_node *new;
new = create_child(parent, false);
if (new == NULL)
return NULL;
if (fill_node(new, cursor) < 0) {
struct callchain_list *call, *tmp;
list_for_each_entry_safe(call, tmp, &new->val, list) {
list_del_init(&call->list);
map__zput(call->ms.map);
free(call);
}
free(new);
return NULL;
}
new->children_hit = 0;
new->hit = period;
new->children_count = 0;
new->count = 1;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
return new;
}
enum match_result {
MATCH_ERROR = -1,
MATCH_EQ,
MATCH_LT,
MATCH_GT,
};
perf report: Fall-back to function name comparison for -g srcline When a callchain entry has no srcline available, we ended up comparing the instruction pointer. I consider this to be not too useful. Rather, I think we should group the entries by function name, which this patch adds. For people who want to split the data on the IP boundary, using `-g address` is the correct choice. Before: ~~~~~ 100.00% 38.86% [.] main | |--61.14%--main inlining.cpp:14 | std::norm<double> complex:664 | std::_Norm_helper<true>::_S_do_it<double> complex:654 | std::abs<double> complex:597 | std::__complex_abs complex:589 | | | |--56.03%--hypot | | | | | |--8.45%--__hypot_finite | | | | | |--7.62%--__hypot_finite | | | | | |--2.29%--__hypot_finite | | | | | |--2.24%--__hypot_finite | | | | | |--2.06%--__hypot_finite | | | | | |--1.81%--__hypot_finite ... ~~~~~ After: ~~~~~ 100.00% 38.86% [.] main | |--61.14%--main inlining.cpp:14 | std::norm<double> complex:664 | std::_Norm_helper<true>::_S_do_it<double> complex:654 | std::abs<double> complex:597 | std::__complex_abs complex:589 | | | |--60.29%--hypot | | | | | --56.03%--__hypot_finite | | | --0.85%--cabs ~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Reviewed-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: Namhyung Kim <namhyung@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20171009203310.17362-7-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-10-10 04:33:00 +08:00
static enum match_result match_chain_strings(const char *left,
const char *right)
perf report: Enable sorting by srcline as key Often it is interesting to know how costly a given source line is in total. Previously, one had to build these sums manually based on all addresses that pointed to the same source line. This patch introduces srcline as a sort key, which will do the aggregation for us. Paired with the recent addition of showing inline frames, this makes perf report much more useful for many C++ work loads. The following shows the new feature in action. First, let's show the status quo output when we sort by address. The result contains many hist entries that generate the same output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g address # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--60.31%--hypot +20 | | | | | |--8.52%--__hypot_finite +273 | | | | | |--7.32%--__hypot_finite +411 ... --35.34%--_start +4194346 __libc_start_main +241 | |--6.65%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--2.70%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--1.69%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ With this patch and `-g srcline` we instead get the following output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g srcline # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--64.02%--hypot | | | | | --59.81%--__hypot_finite | | | --0.53%--cabs | --35.34%--_start __libc_start_main | |--12.48%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20170318214928.9047-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-03-19 05:49:28 +08:00
{
perf report: Don't crash on invalid maps in `-g srcline` mode I just hit a segfault when doing `perf report -g srcline`. Valgrind pointed me at this code as the culprit: ==8359== Invalid read of size 8 ==8359== at 0x3096D9: map__rip_2objdump (map.c:430) ==8359== by 0x2FC1A3: match_chain_srcline (callchain.c:645) ==8359== by 0x2FC1A3: match_chain (callchain.c:700) ==8359== by 0x2FC1A3: append_chain (callchain.c:895) ==8359== by 0x2FC1A3: append_chain_children (callchain.c:846) ==8359== by 0x2FF719: callchain_append (callchain.c:944) ==8359== by 0x2FF719: hist_entry__append_callchain (callchain.c:1058) ==8359== by 0x32FA06: iter_add_single_cumulative_entry (hist.c:908) ==8359== by 0x33195C: hist_entry_iter__add (hist.c:1050) ==8359== by 0x258F65: process_sample_event (builtin-report.c:204) ==8359== by 0x30D60C: perf_session__deliver_event (session.c:1310) ==8359== by 0x30D60C: ordered_events__deliver_event (session.c:119) ==8359== by 0x310D12: __ordered_events__flush (ordered-events.c:210) ==8359== by 0x310D12: ordered_events__flush.part.3 (ordered-events.c:277) ==8359== by 0x30DD3C: perf_session__process_user_event (session.c:1349) ==8359== by 0x30DD3C: perf_session__process_event (session.c:1475) ==8359== by 0x30FC3C: __perf_session__process_events (session.c:1867) ==8359== by 0x30FC3C: perf_session__process_events (session.c:1921) ==8359== by 0x25A985: __cmd_report (builtin-report.c:575) ==8359== by 0x25A985: cmd_report (builtin-report.c:1054) ==8359== by 0x2B9A80: run_builtin (perf.c:296) ==8359== Address 0x70 is not stack'd, malloc'd or (recently) free'd This patch fixes the issue. Signed-off-by: Milian Wolff <milian.wolff@kdab.com> [ Remove dependency from another change ] Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Yao Jin <yao.jin@linux.intel.com> Cc: kernel-team@lge.com Link: http://lkml.kernel.org/r/20170524062129.32529-2-namhyung@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-05-24 14:21:23 +08:00
enum match_result ret = MATCH_EQ;
int cmp;
perf report: Enable sorting by srcline as key Often it is interesting to know how costly a given source line is in total. Previously, one had to build these sums manually based on all addresses that pointed to the same source line. This patch introduces srcline as a sort key, which will do the aggregation for us. Paired with the recent addition of showing inline frames, this makes perf report much more useful for many C++ work loads. The following shows the new feature in action. First, let's show the status quo output when we sort by address. The result contains many hist entries that generate the same output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g address # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--60.31%--hypot +20 | | | | | |--8.52%--__hypot_finite +273 | | | | | |--7.32%--__hypot_finite +411 ... --35.34%--_start +4194346 __libc_start_main +241 | |--6.65%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--2.70%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--1.69%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ With this patch and `-g srcline` we instead get the following output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g srcline # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--64.02%--hypot | | | | | --59.81%--__hypot_finite | | | --0.53%--cabs | --35.34%--_start __libc_start_main | |--12.48%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20170318214928.9047-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-03-19 05:49:28 +08:00
if (left && right)
cmp = strcmp(left, right);
else if (!left && right)
cmp = 1;
else if (left && !right)
cmp = -1;
else
perf report: Fall-back to function name comparison for -g srcline When a callchain entry has no srcline available, we ended up comparing the instruction pointer. I consider this to be not too useful. Rather, I think we should group the entries by function name, which this patch adds. For people who want to split the data on the IP boundary, using `-g address` is the correct choice. Before: ~~~~~ 100.00% 38.86% [.] main | |--61.14%--main inlining.cpp:14 | std::norm<double> complex:664 | std::_Norm_helper<true>::_S_do_it<double> complex:654 | std::abs<double> complex:597 | std::__complex_abs complex:589 | | | |--56.03%--hypot | | | | | |--8.45%--__hypot_finite | | | | | |--7.62%--__hypot_finite | | | | | |--2.29%--__hypot_finite | | | | | |--2.24%--__hypot_finite | | | | | |--2.06%--__hypot_finite | | | | | |--1.81%--__hypot_finite ... ~~~~~ After: ~~~~~ 100.00% 38.86% [.] main | |--61.14%--main inlining.cpp:14 | std::norm<double> complex:664 | std::_Norm_helper<true>::_S_do_it<double> complex:654 | std::abs<double> complex:597 | std::__complex_abs complex:589 | | | |--60.29%--hypot | | | | | --56.03%--__hypot_finite | | | --0.85%--cabs ~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Reviewed-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: Namhyung Kim <namhyung@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20171009203310.17362-7-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-10-10 04:33:00 +08:00
return MATCH_ERROR;
perf report: Enable sorting by srcline as key Often it is interesting to know how costly a given source line is in total. Previously, one had to build these sums manually based on all addresses that pointed to the same source line. This patch introduces srcline as a sort key, which will do the aggregation for us. Paired with the recent addition of showing inline frames, this makes perf report much more useful for many C++ work loads. The following shows the new feature in action. First, let's show the status quo output when we sort by address. The result contains many hist entries that generate the same output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g address # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--60.31%--hypot +20 | | | | | |--8.52%--__hypot_finite +273 | | | | | |--7.32%--__hypot_finite +411 ... --35.34%--_start +4194346 __libc_start_main +241 | |--6.65%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--2.70%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--1.69%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ With this patch and `-g srcline` we instead get the following output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g srcline # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--64.02%--hypot | | | | | --59.81%--__hypot_finite | | | --0.53%--cabs | --35.34%--_start __libc_start_main | |--12.48%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20170318214928.9047-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-03-19 05:49:28 +08:00
if (cmp != 0)
ret = cmp < 0 ? MATCH_LT : MATCH_GT;
return ret;
}
/*
* We need to always use relative addresses because we're aggregating
* callchains from multiple threads, i.e. different address spaces, so
* comparing absolute addresses make no sense as a symbol in a DSO may end up
* in a different address when used in a different binary or even the same
* binary but with some sort of address randomization technique, thus we need
* to compare just relative addresses. -acme
*/
static enum match_result match_chain_dso_addresses(struct map *left_map, u64 left_ip,
struct map *right_map, u64 right_ip)
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
{
struct dso *left_dso = left_map ? left_map->dso : NULL;
struct dso *right_dso = right_map ? right_map->dso : NULL;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
if (left_dso != right_dso)
return left_dso < right_dso ? MATCH_LT : MATCH_GT;
perf report: Fall-back to function name comparison for -g srcline When a callchain entry has no srcline available, we ended up comparing the instruction pointer. I consider this to be not too useful. Rather, I think we should group the entries by function name, which this patch adds. For people who want to split the data on the IP boundary, using `-g address` is the correct choice. Before: ~~~~~ 100.00% 38.86% [.] main | |--61.14%--main inlining.cpp:14 | std::norm<double> complex:664 | std::_Norm_helper<true>::_S_do_it<double> complex:654 | std::abs<double> complex:597 | std::__complex_abs complex:589 | | | |--56.03%--hypot | | | | | |--8.45%--__hypot_finite | | | | | |--7.62%--__hypot_finite | | | | | |--2.29%--__hypot_finite | | | | | |--2.24%--__hypot_finite | | | | | |--2.06%--__hypot_finite | | | | | |--1.81%--__hypot_finite ... ~~~~~ After: ~~~~~ 100.00% 38.86% [.] main | |--61.14%--main inlining.cpp:14 | std::norm<double> complex:664 | std::_Norm_helper<true>::_S_do_it<double> complex:654 | std::abs<double> complex:597 | std::__complex_abs complex:589 | | | |--60.29%--hypot | | | | | --56.03%--__hypot_finite | | | --0.85%--cabs ~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Reviewed-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: Namhyung Kim <namhyung@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20171009203310.17362-7-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-10-10 04:33:00 +08:00
if (left_ip != right_ip)
return left_ip < right_ip ? MATCH_LT : MATCH_GT;
perf report: Enable sorting by srcline as key Often it is interesting to know how costly a given source line is in total. Previously, one had to build these sums manually based on all addresses that pointed to the same source line. This patch introduces srcline as a sort key, which will do the aggregation for us. Paired with the recent addition of showing inline frames, this makes perf report much more useful for many C++ work loads. The following shows the new feature in action. First, let's show the status quo output when we sort by address. The result contains many hist entries that generate the same output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g address # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--60.31%--hypot +20 | | | | | |--8.52%--__hypot_finite +273 | | | | | |--7.32%--__hypot_finite +411 ... --35.34%--_start +4194346 __libc_start_main +241 | |--6.65%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--2.70%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--1.69%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ With this patch and `-g srcline` we instead get the following output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g srcline # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--64.02%--hypot | | | | | --59.81%--__hypot_finite | | | --0.53%--cabs | --35.34%--_start __libc_start_main | |--12.48%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20170318214928.9047-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-03-19 05:49:28 +08:00
return MATCH_EQ;
}
perf report: Enable sorting by srcline as key Often it is interesting to know how costly a given source line is in total. Previously, one had to build these sums manually based on all addresses that pointed to the same source line. This patch introduces srcline as a sort key, which will do the aggregation for us. Paired with the recent addition of showing inline frames, this makes perf report much more useful for many C++ work loads. The following shows the new feature in action. First, let's show the status quo output when we sort by address. The result contains many hist entries that generate the same output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g address # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--60.31%--hypot +20 | | | | | |--8.52%--__hypot_finite +273 | | | | | |--7.32%--__hypot_finite +411 ... --35.34%--_start +4194346 __libc_start_main +241 | |--6.65%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--2.70%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--1.69%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ With this patch and `-g srcline` we instead get the following output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g srcline # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--64.02%--hypot | | | | | --59.81%--__hypot_finite | | | --0.53%--cabs | --35.34%--_start __libc_start_main | |--12.48%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20170318214928.9047-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-03-19 05:49:28 +08:00
static enum match_result match_chain(struct callchain_cursor_node *node,
struct callchain_list *cnode)
{
enum match_result match = MATCH_ERROR;
perf callchain: Compare symbol name for inlined frames when matching The fake symbols we create for inlined frames will represent different functions but can use the symbol start address. This leads to issues when different inline branches all lead to the same function. Before: ~~~~~ $ perf report -s sym -i perf.inlining.data --inline --stdio -g function ... --38.86%--_start __libc_start_main main | --37.57%--std::norm<double> (inlined) std::_Norm_helper<true>::_S_do_it<double> (inlined) | --36.36%--std::abs<double> (inlined) std::__complex_abs (inlined) | --12.24%--std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>::operator() (inlined) std::__detail::__mod<unsigned long, 2147483647ul, 16807ul, 0ul> (inlined) std::__detail::_Mod<unsigned long, 2147483647ul, 16807ul, 0ul, true, true>::__calc (inlined) ~~~~~ Note that this backtrace representation is completely bogus. Complex abs does not call the linear congruential engine! It is just a side-effect of a longer inlined stack being appended to a shorter, different inlined stack, both of which originate in the same function (main). This patch fixes the issue: ~~~~~ $ perf report -s sym -i perf.inlining.data --inline --stdio -g function ... --38.86%--_start __libc_start_main main | |--35.59%--std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inlined) | std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inlined) | | | --34.37%--std::__detail::_Adaptor<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>, double>::operator() (inlined) | std::generate_canonical<double, 53ul, std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inlined) | | | --12.24%--std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>::operator() (inlined) | std::__detail::__mod<unsigned long, 2147483647ul, 16807ul, 0ul> (inlined) | std::__detail::_Mod<unsigned long, 2147483647ul, 16807ul, 0ul, true, true>::__calc (inlined) | --1.99%--std::norm<double> (inlined) std::_Norm_helper<true>::_S_do_it<double> (inlined) std::abs<double> (inlined) std::__complex_abs (inlined) ~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Reviewed-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: Namhyung Kim <namhyung@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20171009203310.17362-10-milian.wolff@kdab.com Cc: Arnaldo Carvalho de Melo <acme@redhat.com> [ Fix up conflict with c1fbc0cf81f1 ("perf callchain: Compare dsos (as well) for CCKEY_FUNCTION"), remove unneeded hunk ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-10-10 04:33:03 +08:00
switch (callchain_param.key) {
case CCKEY_SRCLINE:
match = match_chain_strings(cnode->srcline, node->srcline);
if (match != MATCH_ERROR)
break;
/* otherwise fall-back to symbol-based comparison below */
__fallthrough;
case CCKEY_FUNCTION:
if (node->ms.sym && cnode->ms.sym) {
/*
* Compare inlined frames based on their symbol name
* because different inlined frames will have the same
* symbol start. Otherwise do a faster comparison based
* on the symbol start address.
*/
if (cnode->ms.sym->inlined || node->ms.sym->inlined) {
match = match_chain_strings(cnode->ms.sym->name,
node->ms.sym->name);
if (match != MATCH_ERROR)
break;
} else {
match = match_chain_dso_addresses(cnode->ms.map, cnode->ms.sym->start,
node->ms.map, node->ms.sym->start);
break;
}
}
/* otherwise fall-back to IP-based comparison below */
__fallthrough;
case CCKEY_ADDRESS:
default:
match = match_chain_dso_addresses(cnode->ms.map, cnode->ip, node->ms.map, node->ip);
break;
}
if (match == MATCH_EQ && node->branch) {
cnode->branch_count++;
if (node->branch_from) {
/*
* It's "to" of a branch
*/
cnode->brtype_stat.branch_to = true;
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
if (node->branch_flags.predicted)
cnode->predicted_count++;
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
if (node->branch_flags.abort)
cnode->abort_count++;
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
branch_type_count(&cnode->brtype_stat,
&node->branch_flags,
node->branch_from,
node->ip);
} else {
/*
* It's "from" of a branch
*/
cnode->brtype_stat.branch_to = false;
cnode->cycles_count += node->branch_flags.cycles;
cnode->iter_count += node->nr_loop_iter;
cnode->iter_cycles += node->iter_cycles;
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
cnode->from_count++;
}
}
return match;
}
/*
* Split the parent in two parts (a new child is created) and
* give a part of its callchain to the created child.
* Then create another child to host the given callchain of new branch
*/
static int
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
split_add_child(struct callchain_node *parent,
struct callchain_cursor *cursor,
struct callchain_list *to_split,
u64 idx_parents, u64 idx_local, u64 period)
{
struct callchain_node *new;
struct list_head *old_tail;
unsigned int idx_total = idx_parents + idx_local;
/* split */
new = create_child(parent, true);
if (new == NULL)
return -1;
/* split the callchain and move a part to the new child */
old_tail = parent->val.prev;
list_del_range(&to_split->list, old_tail);
new->val.next = &to_split->list;
new->val.prev = old_tail;
to_split->list.prev = &new->val;
old_tail->next = &new->val;
/* split the hits */
new->hit = parent->hit;
new->children_hit = parent->children_hit;
parent->children_hit = callchain_cumul_hits(new);
new->val_nr = parent->val_nr - idx_local;
parent->val_nr = idx_local;
new->count = parent->count;
new->children_count = parent->children_count;
parent->children_count = callchain_cumul_counts(new);
/* create a new child for the new branch if any */
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (idx_total < cursor->nr) {
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
struct callchain_node *first;
struct callchain_list *cnode;
struct callchain_cursor_node *node;
struct rb_node *p, **pp;
parent->hit = 0;
parent->children_hit += period;
parent->count = 0;
parent->children_count += 1;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
node = callchain_cursor_current(cursor);
new = add_child(parent, cursor, period);
if (new == NULL)
return -1;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
/*
* This is second child since we moved parent's children
* to new (first) child above.
*/
p = parent->rb_root_in.rb_node;
first = rb_entry(p, struct callchain_node, rb_node_in);
cnode = list_first_entry(&first->val, struct callchain_list,
list);
if (match_chain(node, cnode) == MATCH_LT)
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
pp = &p->rb_left;
else
pp = &p->rb_right;
rb_link_node(&new->rb_node_in, p, pp);
rb_insert_color(&new->rb_node_in, &parent->rb_root_in);
} else {
parent->hit = period;
parent->count = 1;
}
return 0;
}
static enum match_result
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
append_chain(struct callchain_node *root,
struct callchain_cursor *cursor,
u64 period);
static int
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
append_chain_children(struct callchain_node *root,
struct callchain_cursor *cursor,
u64 period)
{
struct callchain_node *rnode;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
struct callchain_cursor_node *node;
struct rb_node **p = &root->rb_root_in.rb_node;
struct rb_node *parent = NULL;
node = callchain_cursor_current(cursor);
if (!node)
return -1;
/* lookup in children */
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
while (*p) {
enum match_result ret;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
parent = *p;
rnode = rb_entry(parent, struct callchain_node, rb_node_in);
/* If at least first entry matches, rely to children */
ret = append_chain(rnode, cursor, period);
if (ret == MATCH_EQ)
goto inc_children_hit;
if (ret == MATCH_ERROR)
return -1;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
if (ret == MATCH_LT)
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
p = &parent->rb_left;
else
p = &parent->rb_right;
}
/* nothing in children, add to the current node */
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
rnode = add_child(root, cursor, period);
if (rnode == NULL)
return -1;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
rb_link_node(&rnode->rb_node_in, parent, p);
rb_insert_color(&rnode->rb_node_in, &root->rb_root_in);
inc_children_hit:
root->children_hit += period;
root->children_count++;
return 0;
}
static enum match_result
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
append_chain(struct callchain_node *root,
struct callchain_cursor *cursor,
u64 period)
{
struct callchain_list *cnode;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
u64 start = cursor->pos;
bool found = false;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
u64 matches;
enum match_result cmp = MATCH_ERROR;
/*
* Lookup in the current node
* If we have a symbol, then compare the start to match
* anywhere inside a function, unless function
* mode is disabled.
*/
list_for_each_entry(cnode, &root->val, list) {
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
struct callchain_cursor_node *node;
perf: Fix orphan callchain branches Callchains have markers inside their capture to tell we enter a context (kernel, user, ...). Those are not displayed in the callchains but they are incidentally an active part of the radix tree where callchains are stored, just like any other address. If we have the two following callchains: addr1 -> addr2 -> user context -> addr3 addr1 -> addr2 -> user context -> addr4 addr1 -> addr2 -> addr 5 This is pretty common if addr1 and addr2 are part of an interrupt path, addr3 and addr4 are user addresses and addr5 is a kernel non interrupt path. This will be stored as follows in the tree: addr1 addr2 / \ / addr5 user context / \ addr3 addr4 But we ignore the context markers in the report, hence the addr3 and addr4 will appear as orphan branches: |--28.30%-- hrtimer_interrupt | smp_apic_timer_interrupt | apic_timer_interrupt | | <------------- here, no parent! | | | | | |--11.11%-- 0x7fae7bccb875 | | | | | |--11.11%-- 0xffffffffff60013b | | | | | |--11.11%-- __pthread_mutex_lock_internal | | | | | |--11.11%-- __errno_location Fix this by removing the context markers when we process the callchains to the tree. Reported-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1269274173-20328-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-23 00:09:33 +08:00
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
node = callchain_cursor_current(cursor);
if (!node)
break;
perf: Fix orphan callchain branches Callchains have markers inside their capture to tell we enter a context (kernel, user, ...). Those are not displayed in the callchains but they are incidentally an active part of the radix tree where callchains are stored, just like any other address. If we have the two following callchains: addr1 -> addr2 -> user context -> addr3 addr1 -> addr2 -> user context -> addr4 addr1 -> addr2 -> addr 5 This is pretty common if addr1 and addr2 are part of an interrupt path, addr3 and addr4 are user addresses and addr5 is a kernel non interrupt path. This will be stored as follows in the tree: addr1 addr2 / \ / addr5 user context / \ addr3 addr4 But we ignore the context markers in the report, hence the addr3 and addr4 will appear as orphan branches: |--28.30%-- hrtimer_interrupt | smp_apic_timer_interrupt | apic_timer_interrupt | | <------------- here, no parent! | | | | | |--11.11%-- 0x7fae7bccb875 | | | | | |--11.11%-- 0xffffffffff60013b | | | | | |--11.11%-- __pthread_mutex_lock_internal | | | | | |--11.11%-- __errno_location Fix this by removing the context markers when we process the callchains to the tree. Reported-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1269274173-20328-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-23 00:09:33 +08:00
cmp = match_chain(node, cnode);
if (cmp != MATCH_EQ)
break;
perf: Fix orphan callchain branches Callchains have markers inside their capture to tell we enter a context (kernel, user, ...). Those are not displayed in the callchains but they are incidentally an active part of the radix tree where callchains are stored, just like any other address. If we have the two following callchains: addr1 -> addr2 -> user context -> addr3 addr1 -> addr2 -> user context -> addr4 addr1 -> addr2 -> addr 5 This is pretty common if addr1 and addr2 are part of an interrupt path, addr3 and addr4 are user addresses and addr5 is a kernel non interrupt path. This will be stored as follows in the tree: addr1 addr2 / \ / addr5 user context / \ addr3 addr4 But we ignore the context markers in the report, hence the addr3 and addr4 will appear as orphan branches: |--28.30%-- hrtimer_interrupt | smp_apic_timer_interrupt | apic_timer_interrupt | | <------------- here, no parent! | | | | | |--11.11%-- 0x7fae7bccb875 | | | | | |--11.11%-- 0xffffffffff60013b | | | | | |--11.11%-- __pthread_mutex_lock_internal | | | | | |--11.11%-- __errno_location Fix this by removing the context markers when we process the callchains to the tree. Reported-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1269274173-20328-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-23 00:09:33 +08:00
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
found = true;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
callchain_cursor_advance(cursor);
}
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
/* matches not, relay no the parent */
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (!found) {
WARN_ONCE(cmp == MATCH_ERROR, "Chain comparison error\n");
return cmp;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
}
matches = cursor->pos - start;
/* we match only a part of the node. Split it and add the new chain */
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (matches < root->val_nr) {
if (split_add_child(root, cursor, cnode, start, matches,
period) < 0)
return MATCH_ERROR;
return MATCH_EQ;
}
/* we match 100% of the path, increment the hit */
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (matches == root->val_nr && cursor->pos == cursor->nr) {
root->hit += period;
root->count++;
return MATCH_EQ;
}
/* We match the node and still have a part remaining */
if (append_chain_children(root, cursor, period) < 0)
return MATCH_ERROR;
return MATCH_EQ;
}
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
int callchain_append(struct callchain_root *root,
struct callchain_cursor *cursor,
u64 period)
{
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (!cursor->nr)
perf: Fix orphan callchain branches Callchains have markers inside their capture to tell we enter a context (kernel, user, ...). Those are not displayed in the callchains but they are incidentally an active part of the radix tree where callchains are stored, just like any other address. If we have the two following callchains: addr1 -> addr2 -> user context -> addr3 addr1 -> addr2 -> user context -> addr4 addr1 -> addr2 -> addr 5 This is pretty common if addr1 and addr2 are part of an interrupt path, addr3 and addr4 are user addresses and addr5 is a kernel non interrupt path. This will be stored as follows in the tree: addr1 addr2 / \ / addr5 user context / \ addr3 addr4 But we ignore the context markers in the report, hence the addr3 and addr4 will appear as orphan branches: |--28.30%-- hrtimer_interrupt | smp_apic_timer_interrupt | apic_timer_interrupt | | <------------- here, no parent! | | | | | |--11.11%-- 0x7fae7bccb875 | | | | | |--11.11%-- 0xffffffffff60013b | | | | | |--11.11%-- __pthread_mutex_lock_internal | | | | | |--11.11%-- __errno_location Fix this by removing the context markers when we process the callchains to the tree. Reported-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1269274173-20328-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-23 00:09:33 +08:00
return 0;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
callchain_cursor_commit(cursor);
perf: Fix orphan callchain branches Callchains have markers inside their capture to tell we enter a context (kernel, user, ...). Those are not displayed in the callchains but they are incidentally an active part of the radix tree where callchains are stored, just like any other address. If we have the two following callchains: addr1 -> addr2 -> user context -> addr3 addr1 -> addr2 -> user context -> addr4 addr1 -> addr2 -> addr 5 This is pretty common if addr1 and addr2 are part of an interrupt path, addr3 and addr4 are user addresses and addr5 is a kernel non interrupt path. This will be stored as follows in the tree: addr1 addr2 / \ / addr5 user context / \ addr3 addr4 But we ignore the context markers in the report, hence the addr3 and addr4 will appear as orphan branches: |--28.30%-- hrtimer_interrupt | smp_apic_timer_interrupt | apic_timer_interrupt | | <------------- here, no parent! | | | | | |--11.11%-- 0x7fae7bccb875 | | | | | |--11.11%-- 0xffffffffff60013b | | | | | |--11.11%-- __pthread_mutex_lock_internal | | | | | |--11.11%-- __errno_location Fix this by removing the context markers when we process the callchains to the tree. Reported-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1269274173-20328-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-23 00:09:33 +08:00
if (append_chain_children(&root->node, cursor, period) < 0)
return -1;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (cursor->nr > root->max_depth)
root->max_depth = cursor->nr;
perf: Fix orphan callchain branches Callchains have markers inside their capture to tell we enter a context (kernel, user, ...). Those are not displayed in the callchains but they are incidentally an active part of the radix tree where callchains are stored, just like any other address. If we have the two following callchains: addr1 -> addr2 -> user context -> addr3 addr1 -> addr2 -> user context -> addr4 addr1 -> addr2 -> addr 5 This is pretty common if addr1 and addr2 are part of an interrupt path, addr3 and addr4 are user addresses and addr5 is a kernel non interrupt path. This will be stored as follows in the tree: addr1 addr2 / \ / addr5 user context / \ addr3 addr4 But we ignore the context markers in the report, hence the addr3 and addr4 will appear as orphan branches: |--28.30%-- hrtimer_interrupt | smp_apic_timer_interrupt | apic_timer_interrupt | | <------------- here, no parent! | | | | | |--11.11%-- 0x7fae7bccb875 | | | | | |--11.11%-- 0xffffffffff60013b | | | | | |--11.11%-- __pthread_mutex_lock_internal | | | | | |--11.11%-- __errno_location Fix this by removing the context markers when we process the callchains to the tree. Reported-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <1269274173-20328-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-23 00:09:33 +08:00
return 0;
}
static int
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
merge_chain_branch(struct callchain_cursor *cursor,
struct callchain_node *dst, struct callchain_node *src)
{
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
struct callchain_cursor_node **old_last = cursor->last;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
struct callchain_node *child;
struct callchain_list *list, *next_list;
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
struct rb_node *n;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
int old_pos = cursor->nr;
int err = 0;
list_for_each_entry_safe(list, next_list, &src->val, list) {
callchain_cursor_append(cursor, list->ip, &list->ms,
false, NULL, 0, 0, 0, list->srcline);
list_del_init(&list->list);
map__zput(list->ms.map);
free(list);
}
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (src->hit) {
callchain_cursor_commit(cursor);
if (append_chain_children(dst, cursor, src->hit) < 0)
return -1;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
}
perf callchain: Convert children list to rbtree Current collapse stage has a scalability problem which can be reproduced easily with a parallel kernel build. This is because it needs to traverse every children of callchains linearly during the collapse/merge stage. Converting it to a rbtree reduced the overhead significantly. On my 400MB perf.data file which recorded with make -j32 kernel build: $ time perf --no-pager report --stdio > /dev/null before: real 6m22.073s user 6m18.683s sys 0m0.706s after: real 0m20.780s user 0m19.962s sys 0m0.689s During the perf report the overhead on append_chain_children went down from 96.69% to 18.16%: - 18.16% perf perf [.] append_chain_children - append_chain_children - 77.48% append_chain_children + 69.79% merge_chain_branch - 22.96% append_chain_children + 67.44% merge_chain_branch + 30.15% append_chain_children + 2.41% callchain_append + 7.25% callchain_append + 12.26% callchain_append + 10.22% merge_chain_branch + 11.58% perf perf [.] dso__find_symbol + 8.02% perf perf [.] sort__comm_cmp + 5.48% perf libc-2.17.so [.] malloc_consolidate Reported-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1381468543-25334-2-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-11 13:15:36 +08:00
n = rb_first(&src->rb_root_in);
while (n) {
child = container_of(n, struct callchain_node, rb_node_in);
n = rb_next(n);
rb_erase(&child->rb_node_in, &src->rb_root_in);
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
err = merge_chain_branch(cursor, dst, child);
if (err)
break;
free(child);
}
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
cursor->nr = old_pos;
cursor->last = old_last;
return err;
}
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
int callchain_merge(struct callchain_cursor *cursor,
struct callchain_root *dst, struct callchain_root *src)
{
return merge_chain_branch(cursor, &dst->node, &src->node);
}
int callchain_cursor_append(struct callchain_cursor *cursor,
u64 ip, struct map_symbol *ms,
perf report: Add branch flag to callchain cursor node Since the branch ip has been added to call stack for easier browsing, this patch adds more branch information. For example, add a flag to indicate if this ip is a branch, and also add with the branch flag. Then we can know if the cursor node represents a branch and know what the branch flag it has. The branch history code has a loop detection pass that removes loops. It would be nice for knowing how many loops were removed then in next steps, we can compute out the average number of iterations. For example: Before remove_loops(), entry0: from = 0x100, to = 0x200 entry1: from = 0x300, to = 0x250 entry2: from = 0x300, to = 0x250 entry3: from = 0x300, to = 0x250 entry4: from = 0x700, to = 0x800 After remove_loops() entry0: from = 0x100, to = 0x200 entry1: from = 0x300, to = 0x250 entry2: from = 0x700, to = 0x800 The original entry2 and entry3 are removed. So the number of iterations (from = 0x300, to = 0x250) is equal to removed number + 1 (2 + 1). iterations = removed number + 1; average iteractions = Sum(iteractions) / number of samples This formula ignores other cases, for example, iterations cross multiple buffers and one buffer contains 2+ loops. Because in practice, it's good enough. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Acked-by: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Linux-kernel@vger.kernel.org Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/n/1477876794-30749-2-git-send-email-yao.jin@linux.intel.com [ Renamed 'iter' to 'nr_loop_iter' for clarity ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-10-31 09:19:49 +08:00
bool branch, struct branch_flags *flags,
int nr_loop_iter, u64 iter_cycles, u64 branch_from,
const char *srcline)
{
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
struct callchain_cursor_node *node = *cursor->last;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (!node) {
node = calloc(1, sizeof(*node));
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (!node)
return -ENOMEM;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
*cursor->last = node;
}
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
node->ip = ip;
map__zput(node->ms.map);
node->ms = *ms;
map__get(node->ms.map);
perf report: Add branch flag to callchain cursor node Since the branch ip has been added to call stack for easier browsing, this patch adds more branch information. For example, add a flag to indicate if this ip is a branch, and also add with the branch flag. Then we can know if the cursor node represents a branch and know what the branch flag it has. The branch history code has a loop detection pass that removes loops. It would be nice for knowing how many loops were removed then in next steps, we can compute out the average number of iterations. For example: Before remove_loops(), entry0: from = 0x100, to = 0x200 entry1: from = 0x300, to = 0x250 entry2: from = 0x300, to = 0x250 entry3: from = 0x300, to = 0x250 entry4: from = 0x700, to = 0x800 After remove_loops() entry0: from = 0x100, to = 0x200 entry1: from = 0x300, to = 0x250 entry2: from = 0x700, to = 0x800 The original entry2 and entry3 are removed. So the number of iterations (from = 0x300, to = 0x250) is equal to removed number + 1 (2 + 1). iterations = removed number + 1; average iteractions = Sum(iteractions) / number of samples This formula ignores other cases, for example, iterations cross multiple buffers and one buffer contains 2+ loops. Because in practice, it's good enough. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Acked-by: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Linux-kernel@vger.kernel.org Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/n/1477876794-30749-2-git-send-email-yao.jin@linux.intel.com [ Renamed 'iter' to 'nr_loop_iter' for clarity ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-10-31 09:19:49 +08:00
node->branch = branch;
node->nr_loop_iter = nr_loop_iter;
node->iter_cycles = iter_cycles;
node->srcline = srcline;
perf report: Add branch flag to callchain cursor node Since the branch ip has been added to call stack for easier browsing, this patch adds more branch information. For example, add a flag to indicate if this ip is a branch, and also add with the branch flag. Then we can know if the cursor node represents a branch and know what the branch flag it has. The branch history code has a loop detection pass that removes loops. It would be nice for knowing how many loops were removed then in next steps, we can compute out the average number of iterations. For example: Before remove_loops(), entry0: from = 0x100, to = 0x200 entry1: from = 0x300, to = 0x250 entry2: from = 0x300, to = 0x250 entry3: from = 0x300, to = 0x250 entry4: from = 0x700, to = 0x800 After remove_loops() entry0: from = 0x100, to = 0x200 entry1: from = 0x300, to = 0x250 entry2: from = 0x700, to = 0x800 The original entry2 and entry3 are removed. So the number of iterations (from = 0x300, to = 0x250) is equal to removed number + 1 (2 + 1). iterations = removed number + 1; average iteractions = Sum(iteractions) / number of samples This formula ignores other cases, for example, iterations cross multiple buffers and one buffer contains 2+ loops. Because in practice, it's good enough. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Acked-by: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Linux-kernel@vger.kernel.org Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/n/1477876794-30749-2-git-send-email-yao.jin@linux.intel.com [ Renamed 'iter' to 'nr_loop_iter' for clarity ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-10-31 09:19:49 +08:00
if (flags)
memcpy(&node->branch_flags, flags,
sizeof(struct branch_flags));
node->branch_from = branch_from;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
cursor->nr++;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
cursor->last = &node->next;
return 0;
}
int sample__resolve_callchain(struct perf_sample *sample,
struct callchain_cursor *cursor, struct symbol **parent,
struct evsel *evsel, struct addr_location *al,
int max_stack)
{
perf report: Make --branch-history work without callgraphs(-g) option in perf record perf record -b -g <command> perf report --branch-history This merges the LBRs with the callgraphs. However it would be nice if it also works without callgraphs (-g) set in perf record, so that only the LBRs are displayed. But currently perf report errors in this case. For example, perf record -b <command> perf report --branch-history Error: Selected -g or --branch-history but no callchain data. Did you call 'perf record' without -g? This patch displays the LBRs only even if callgraphs(-g) is not enabled in perf record. Change log: v2: According to Milian Wolff's comment, change the obsolete error message. Now the error message is: ┌─Error:─────────────────────────────────────┐ │Selected -g or --branch-history. │ │But no callchain or branch data. │ │Did you call 'perf record' without -g or -b?│ │ │ │ │ │Press any key... │ └────────────────────────────────────────────┘ When passing the last parameter to hists__fprintf, changes "|" to "||". hists__fprintf(hists, !quiet, 0, 0, rep->min_percent, stdout, symbol_conf.use_callchain || symbol_conf.show_branchflag_count); Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1494240182-28899-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-05-08 18:43:02 +08:00
if (sample->callchain == NULL && !symbol_conf.show_branchflag_count)
return 0;
if (symbol_conf.use_callchain || symbol_conf.cumulate_callchain ||
perf report: Make --branch-history work without callgraphs(-g) option in perf record perf record -b -g <command> perf report --branch-history This merges the LBRs with the callgraphs. However it would be nice if it also works without callgraphs (-g) set in perf record, so that only the LBRs are displayed. But currently perf report errors in this case. For example, perf record -b <command> perf report --branch-history Error: Selected -g or --branch-history but no callchain data. Did you call 'perf record' without -g? This patch displays the LBRs only even if callgraphs(-g) is not enabled in perf record. Change log: v2: According to Milian Wolff's comment, change the obsolete error message. Now the error message is: ┌─Error:─────────────────────────────────────┐ │Selected -g or --branch-history. │ │But no callchain or branch data. │ │Did you call 'perf record' without -g or -b?│ │ │ │ │ │Press any key... │ └────────────────────────────────────────────┘ When passing the last parameter to hists__fprintf, changes "|" to "||". hists__fprintf(hists, !quiet, 0, 0, rep->min_percent, stdout, symbol_conf.use_callchain || symbol_conf.show_branchflag_count); Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1494240182-28899-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-05-08 18:43:02 +08:00
perf_hpp_list.parent || symbol_conf.show_branchflag_count) {
return thread__resolve_callchain(al->thread, cursor, evsel, sample,
parent, al, max_stack);
}
return 0;
}
int hist_entry__append_callchain(struct hist_entry *he, struct perf_sample *sample)
{
perf report: Make --branch-history work without callgraphs(-g) option in perf record perf record -b -g <command> perf report --branch-history This merges the LBRs with the callgraphs. However it would be nice if it also works without callgraphs (-g) set in perf record, so that only the LBRs are displayed. But currently perf report errors in this case. For example, perf record -b <command> perf report --branch-history Error: Selected -g or --branch-history but no callchain data. Did you call 'perf record' without -g? This patch displays the LBRs only even if callgraphs(-g) is not enabled in perf record. Change log: v2: According to Milian Wolff's comment, change the obsolete error message. Now the error message is: ┌─Error:─────────────────────────────────────┐ │Selected -g or --branch-history. │ │But no callchain or branch data. │ │Did you call 'perf record' without -g or -b?│ │ │ │ │ │Press any key... │ └────────────────────────────────────────────┘ When passing the last parameter to hists__fprintf, changes "|" to "||". hists__fprintf(hists, !quiet, 0, 0, rep->min_percent, stdout, symbol_conf.use_callchain || symbol_conf.show_branchflag_count); Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1494240182-28899-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-05-08 18:43:02 +08:00
if ((!symbol_conf.use_callchain || sample->callchain == NULL) &&
!symbol_conf.show_branchflag_count)
return 0;
return callchain_append(he->callchain, &callchain_cursor, sample->period);
}
int fill_callchain_info(struct addr_location *al, struct callchain_cursor_node *node,
bool hide_unresolved)
{
al->maps = node->ms.maps;
al->map = node->ms.map;
al->sym = node->ms.sym;
perf report: Use srcline from callchain for hist entries This also removes the symbol name from the srcline column, more on this below. This ensures we use the correct srcline, which could originate from a potentially inlined function. The hist entries used to query for the srcline based purely on the IP, which leads to wrong results for inlined entries. Before: ~~~~~ perf report --inline -s srcline -g none --stdio ... # Children Self Source:Line # ........ ........ .................................................................................................................................. # 94.23% 0.00% __libc_start_main+18446603487898210537 94.23% 0.00% _start+41 44.58% 0.00% main+100 44.58% 0.00% std::_Norm_helper<true>::_S_do_it<double>+100 44.58% 0.00% std::__complex_abs+100 44.58% 0.00% std::abs<double>+100 44.58% 0.00% std::norm<double>+100 36.01% 0.00% hypot+18446603487892193300 25.81% 0.00% main+41 25.81% 0.00% std::__detail::_Adaptor<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>, double>::operator()+41 25.81% 0.00% std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> >+41 25.75% 25.75% random.h:143 18.39% 0.00% main+57 18.39% 0.00% std::__detail::_Adaptor<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>, double>::operator()+57 18.39% 0.00% std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> >+57 13.80% 13.80% random.tcc:3330 5.64% 0.00% ??:0 4.13% 4.13% __hypot_finite+163 4.13% 0.00% __hypot_finite+18446603487892193443 ... ~~~~~ After: ~~~~~ perf report --inline -s srcline -g none --stdio ... # Children Self Source:Line # ........ ........ ........................................... # 94.30% 1.19% main.cpp:39 94.23% 0.00% __libc_start_main+18446603487898210537 94.23% 0.00% _start+41 48.44% 1.70% random.h:1823 48.44% 0.00% random.h:1814 46.74% 2.53% random.h:185 44.68% 0.10% complex:589 44.68% 0.00% complex:597 44.68% 0.00% complex:654 44.68% 0.00% complex:664 40.61% 13.80% random.tcc:3330 36.01% 0.00% hypot+18446603487892193300 26.81% 0.00% random.h:151 26.81% 0.00% random.h:332 25.75% 25.75% random.h:143 5.64% 0.00% ??:0 4.13% 4.13% __hypot_finite+163 4.13% 0.00% __hypot_finite+18446603487892193443 ... ~~~~~ Note that this change removes the symbol from the source:line hist column. If this information is desired, users should explicitly query for it if needed. I.e. run this command instead: ~~~~~ perf report --inline -s sym,srcline -g none --stdio ... # To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 1K of event 'cycles:uppp' # Event count (approx.): 1381229476 # # Children Self Symbol Source:Line # ........ ........ ................................................................................................................................... ........................................... # 94.30% 1.19% [.] main main.cpp:39 94.23% 0.00% [.] __libc_start_main __libc_start_main+18446603487898210537 94.23% 0.00% [.] _start _start+41 48.44% 0.00% [.] std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inlined) random.h:1814 48.44% 0.00% [.] std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inlined) random.h:1823 46.74% 0.00% [.] std::__detail::_Adaptor<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>, double>::operator() (inlined) random.h:185 44.68% 0.00% [.] std::_Norm_helper<true>::_S_do_it<double> (inlined) complex:654 44.68% 0.00% [.] std::__complex_abs (inlined) complex:589 44.68% 0.00% [.] std::abs<double> (inlined) complex:597 44.68% 0.00% [.] std::norm<double> (inlined) complex:664 39.80% 13.59% [.] std::generate_canonical<double, 53ul, std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > random.tcc:3330 36.01% 0.00% [.] hypot hypot+18446603487892193300 26.81% 0.00% [.] std::__detail::__mod<unsigned long, 2147483647ul, 16807ul, 0ul> (inlined) random.h:151 26.81% 0.00% [.] std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>::operator() (inlined) random.h:332 25.75% 0.00% [.] std::__detail::_Mod<unsigned long, 2147483647ul, 16807ul, 0ul, true, true>::__calc (inlined) random.h:143 25.19% 25.19% [.] std::generate_canonical<double, 53ul, std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > random.h:143 4.13% 4.13% [.] __hypot_finite __hypot_finite+163 4.13% 0.00% [.] __hypot_finite __hypot_finite+18446603487892193443 ... ~~~~~ Compared to the old behavior, this reduces duplication in the output. Before we used to print the symbol name in the srcline column even when the sym column was explicitly requested. I.e. the output was: ~~~~~ perf report --inline -s sym,srcline -g none --stdio ... # To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 1K of event 'cycles:uppp' # Event count (approx.): 1381229476 # # Children Self Symbol Source:Line # ........ ........ ................................................................................................................................... .................................................................................................................................. # 94.23% 0.00% [.] __libc_start_main __libc_start_main+18446603487898210537 94.23% 0.00% [.] _start _start+41 44.58% 0.00% [.] main main+100 44.58% 0.00% [.] std::_Norm_helper<true>::_S_do_it<double> (inlined) std::_Norm_helper<true>::_S_do_it<double>+100 44.58% 0.00% [.] std::__complex_abs (inlined) std::__complex_abs+100 44.58% 0.00% [.] std::abs<double> (inlined) std::abs<double>+100 44.58% 0.00% [.] std::norm<double> (inlined) std::norm<double>+100 36.01% 0.00% [.] hypot hypot+18446603487892193300 25.81% 0.00% [.] main main+41 25.81% 0.00% [.] std::__detail::_Adaptor<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>, double>::operator() (inlined) std::__detail::_Adaptor<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>, double>::operator()+41 25.81% 0.00% [.] std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inlined) std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> >+41 25.69% 25.69% [.] std::generate_canonical<double, 53ul, std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > random.h:143 18.39% 0.00% [.] main main+57 18.39% 0.00% [.] std::__detail::_Adaptor<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>, double>::operator() (inlined) std::__detail::_Adaptor<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul>, double>::operator()+57 18.39% 0.00% [.] std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inlined) std::uniform_real_distribution<double>::operator()<std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> >+57 13.80% 13.80% [.] std::generate_canonical<double, 53ul, std::linear_congruential_engine<unsigned long, 16807ul, 0ul, 2147483647ul> > random.tcc:3330 4.13% 4.13% [.] __hypot_finite __hypot_finite+163 4.13% 0.00% [.] __hypot_finite __hypot_finite+18446603487892193443 ... ~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jin Yao <yao.jin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20171019113836.5548-5-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-10-19 19:38:35 +08:00
al->srcline = node->srcline;
al->addr = node->ip;
if (al->sym == NULL) {
if (hide_unresolved)
return 0;
if (al->map == NULL)
goto out;
}
perf maps: Use a pointer for kmaps struct maps is reference counted, using a pointer is more idiomatic. Committer notes: Delay: maps = machine__kernel_maps(&vmlinux); To after: machine__init(&vmlinux, "", HOST_KERNEL_ID); To avoid this on f34: In file included from /var/home/acme/git/perf/tools/perf/util/build-id.h:10, from /var/home/acme/git/perf/tools/perf/util/dso.h:13, from tests/vmlinux-kallsyms.c:8: In function ‘machine__kernel_maps’, inlined from ‘test__vmlinux_matches_kallsyms’ at tests/vmlinux-kallsyms.c:122:22: /var/home/acme/git/perf/tools/perf/util/machine.h:86:23: error: ‘vmlinux.kmaps’ is used uninitialized [-Werror=uninitialized] 86 | return machine->kmaps; | ~~~~~~~^~~~~~~ tests/vmlinux-kallsyms.c: In function ‘test__vmlinux_matches_kallsyms’: tests/vmlinux-kallsyms.c:121:34: note: ‘vmlinux’ declared here 121 | struct machine kallsyms, vmlinux; | ^~~~~~~ cc1: all warnings being treated as errors Signed-off-by: Ian Rogers <irogers@google.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Bayduraev <alexey.v.bayduraev@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: André Almeida <andrealmeid@collabora.com> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Darren Hart <dvhart@infradead.org> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Dmitriy Vyukov <dvyukov@google.com> Cc: Eric Dumazet <edumazet@google.com> Cc: German Gomez <german.gomez@arm.com> Cc: Hao Luo <haoluo@google.com> Cc: James Clark <james.clark@arm.com> Cc: Jin Yao <yao.jin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: John Garry <john.garry@huawei.com> Cc: Kajol Jain <kjain@linux.ibm.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Leo Yan <leo.yan@linaro.org> Cc: Madhavan Srinivasan <maddy@linux.ibm.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Miaoqian Lin <linmq006@gmail.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Riccardo Mancini <rickyman7@gmail.com> Cc: Shunsuke Nakamura <nakamura.shun@fujitsu.com> Cc: Song Liu <song@kernel.org> Cc: Stephane Eranian <eranian@google.com> Cc: Stephen Brennan <stephen.s.brennan@oracle.com> Cc: Steven Rostedt (VMware) <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Thomas Richter <tmricht@linux.ibm.com> Cc: Yury Norov <yury.norov@gmail.com> Link: http://lore.kernel.org/lkml/20220211103415.2737789-6-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2022-02-11 18:33:58 +08:00
if (al->maps == machine__kernel_maps(al->maps->machine)) {
if (machine__is_host(al->maps->machine)) {
al->cpumode = PERF_RECORD_MISC_KERNEL;
al->level = 'k';
} else {
al->cpumode = PERF_RECORD_MISC_GUEST_KERNEL;
al->level = 'g';
}
} else {
if (machine__is_host(al->maps->machine)) {
al->cpumode = PERF_RECORD_MISC_USER;
al->level = '.';
} else if (perf_guest) {
al->cpumode = PERF_RECORD_MISC_GUEST_USER;
al->level = 'u';
} else {
al->cpumode = PERF_RECORD_MISC_HYPERVISOR;
al->level = 'H';
}
}
out:
return 1;
}
char *callchain_list__sym_name(struct callchain_list *cl,
char *bf, size_t bfsize, bool show_dso)
{
perf report: Enable sorting by srcline as key Often it is interesting to know how costly a given source line is in total. Previously, one had to build these sums manually based on all addresses that pointed to the same source line. This patch introduces srcline as a sort key, which will do the aggregation for us. Paired with the recent addition of showing inline frames, this makes perf report much more useful for many C++ work loads. The following shows the new feature in action. First, let's show the status quo output when we sort by address. The result contains many hist entries that generate the same output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g address # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--60.31%--hypot +20 | | | | | |--8.52%--__hypot_finite +273 | | | | | |--7.32%--__hypot_finite +411 ... --35.34%--_start +4194346 __libc_start_main +241 | |--6.65%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--2.70%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) | |--1.69%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ With this patch and `-g srcline` we instead get the following output: ~~~~~~~~~~~~~~~~ $ perf report --stdio --inline -g srcline # Children Self Command Shared Object Symbol # ........ ........ ............ ................... ......................................... # 99.89% 35.34% cpp-inlining cpp-inlining [.] main | |--64.55%--main complex:655 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/complex:664 (inline) | | | |--64.02%--hypot | | | | | --59.81%--__hypot_finite | | | --0.53%--cabs | --35.34%--_start __libc_start_main | |--12.48%--main random.tcc:3326 | /home/milian/projects/kdab/rnd/hotspot/tests/test-clients/cpp-inlining/main.cpp:39 (inline) | /usr/include/c++/6.3.1/bits/random.h:1809 (inline) | /usr/include/c++/6.3.1/bits/random.h:1818 (inline) | /usr/include/c++/6.3.1/bits/random.h:185 (inline) ... ~~~~~~~~~~~~~~~~ Signed-off-by: Milian Wolff <milian.wolff@kdab.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Yao Jin <yao.jin@linux.intel.com> Link: http://lkml.kernel.org/r/20170318214928.9047-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-03-19 05:49:28 +08:00
bool show_addr = callchain_param.key == CCKEY_ADDRESS;
bool show_srcline = show_addr || callchain_param.key == CCKEY_SRCLINE;
int printed;
if (cl->ms.sym) {
const char *inlined = cl->ms.sym->inlined ? " (inlined)" : "";
if (show_srcline && cl->srcline)
printed = scnprintf(bf, bfsize, "%s %s%s",
cl->ms.sym->name, cl->srcline,
inlined);
else
printed = scnprintf(bf, bfsize, "%s%s",
cl->ms.sym->name, inlined);
} else
printed = scnprintf(bf, bfsize, "%#" PRIx64, cl->ip);
if (show_dso)
scnprintf(bf + printed, bfsize - printed, " %s",
cl->ms.map ?
cl->ms.map->dso->short_name :
"unknown");
return bf;
}
char *callchain_node__scnprintf_value(struct callchain_node *node,
char *bf, size_t bfsize, u64 total)
{
double percent = 0.0;
u64 period = callchain_cumul_hits(node);
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
unsigned count = callchain_cumul_counts(node);
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
if (callchain_param.mode == CHAIN_FOLDED) {
period = node->hit;
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
count = node->count;
}
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
switch (callchain_param.value) {
case CCVAL_PERIOD:
scnprintf(bf, bfsize, "%"PRIu64, period);
break;
case CCVAL_COUNT:
scnprintf(bf, bfsize, "%u", count);
break;
case CCVAL_PERCENT:
default:
if (total)
percent = period * 100.0 / total;
scnprintf(bf, bfsize, "%.2f%%", percent);
break;
}
return bf;
}
int callchain_node__fprintf_value(struct callchain_node *node,
FILE *fp, u64 total)
{
double percent = 0.0;
u64 period = callchain_cumul_hits(node);
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
unsigned count = callchain_cumul_counts(node);
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
if (callchain_param.mode == CHAIN_FOLDED) {
period = node->hit;
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
count = node->count;
}
perf report: Add callchain value option Now -g/--call-graph option supports how to display callchain values. Possible values are 'percent', 'period' and 'count'. The percent is same as before and it's the default behavior. The period displays the raw period value rather than the percentage. The count displays the number of occurrences. $ perf report --no-children --stdio -g percent ... 39.93% swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--28.63%-- start_secondary | --11.30%-- rest_init $ perf report --no-children --show-total-period --stdio -g period ... 39.93% 13018705 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--9334403-- start_secondary | --3684302-- rest_init $ perf report --no-children --show-nr-samples --stdio -g count ... 39.93% 80 swapper [kernel.vmlinux] [k] intel_idel | ---intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry | |--57-- start_secondary | --23-- rest_init Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-6-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:41 +08:00
switch (callchain_param.value) {
case CCVAL_PERIOD:
return fprintf(fp, "%"PRIu64, period);
case CCVAL_COUNT:
return fprintf(fp, "%u", count);
case CCVAL_PERCENT:
default:
if (total)
percent = period * 100.0 / total;
return percent_color_fprintf(fp, "%.2f%%", percent);
}
return 0;
}
static void callchain_counts_value(struct callchain_node *node,
u64 *branch_count, u64 *predicted_count,
u64 *abort_count, u64 *cycles_count)
{
struct callchain_list *clist;
list_for_each_entry(clist, &node->val, list) {
if (branch_count)
*branch_count += clist->branch_count;
if (predicted_count)
*predicted_count += clist->predicted_count;
if (abort_count)
*abort_count += clist->abort_count;
if (cycles_count)
*cycles_count += clist->cycles_count;
}
}
static int callchain_node_branch_counts_cumul(struct callchain_node *node,
u64 *branch_count,
u64 *predicted_count,
u64 *abort_count,
u64 *cycles_count)
{
struct callchain_node *child;
struct rb_node *n;
n = rb_first(&node->rb_root_in);
while (n) {
child = rb_entry(n, struct callchain_node, rb_node_in);
n = rb_next(n);
callchain_node_branch_counts_cumul(child, branch_count,
predicted_count,
abort_count,
cycles_count);
callchain_counts_value(child, branch_count,
predicted_count, abort_count,
cycles_count);
}
return 0;
}
int callchain_branch_counts(struct callchain_root *root,
u64 *branch_count, u64 *predicted_count,
u64 *abort_count, u64 *cycles_count)
{
if (branch_count)
*branch_count = 0;
if (predicted_count)
*predicted_count = 0;
if (abort_count)
*abort_count = 0;
if (cycles_count)
*cycles_count = 0;
return callchain_node_branch_counts_cumul(&root->node,
branch_count,
predicted_count,
abort_count,
cycles_count);
}
2017-07-18 20:13:12 +08:00
static int count_pri64_printf(int idx, const char *str, u64 value, char *bf, int bfsize)
{
int printed;
printed = scnprintf(bf, bfsize, "%s%s:%" PRId64 "", (idx) ? " " : " (", str, value);
return printed;
}
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
static int count_float_printf(int idx, const char *str, float value,
char *bf, int bfsize, float threshold)
2017-07-18 20:13:12 +08:00
{
int printed;
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
if (threshold != 0.0 && value < threshold)
return 0;
2017-07-18 20:13:12 +08:00
printed = scnprintf(bf, bfsize, "%s%s:%.1f%%", (idx) ? " " : " (", str, value);
return printed;
}
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
static int branch_to_str(char *bf, int bfsize,
u64 branch_count, u64 predicted_count,
u64 abort_count,
struct branch_type_stat *brtype_stat)
{
int printed, i = 0;
printed = branch_type_str(brtype_stat, bf, bfsize);
if (printed)
i++;
2017-07-18 20:13:12 +08:00
if (predicted_count < branch_count) {
printed += count_float_printf(i++, "predicted",
predicted_count * 100.0 / branch_count,
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
bf + printed, bfsize - printed, 0.0);
2017-07-18 20:13:12 +08:00
}
2017-07-18 20:13:12 +08:00
if (abort_count) {
printed += count_float_printf(i++, "abort",
abort_count * 100.0 / branch_count,
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
bf + printed, bfsize - printed, 0.1);
}
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
if (i)
printed += scnprintf(bf + printed, bfsize - printed, ")");
return printed;
}
static int branch_from_str(char *bf, int bfsize,
u64 branch_count,
u64 cycles_count, u64 iter_count,
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
u64 iter_cycles, u64 from_count)
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
{
int printed = 0, i = 0;
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
u64 cycles, v = 0;
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
2017-07-18 20:13:12 +08:00
cycles = cycles_count / branch_count;
if (cycles) {
printed += count_pri64_printf(i++, "cycles",
cycles,
bf + printed, bfsize - printed);
}
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
if (iter_count && from_count) {
v = iter_count / from_count;
if (v) {
printed += count_pri64_printf(i++, "iter",
v, bf + printed, bfsize - printed);
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
printed += count_pri64_printf(i++, "avg_cycles",
iter_cycles / iter_count,
bf + printed, bfsize - printed);
}
}
2017-07-18 20:13:12 +08:00
if (i)
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
printed += scnprintf(bf + printed, bfsize - printed, ")");
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
return printed;
}
static int counts_str_build(char *bf, int bfsize,
u64 branch_count, u64 predicted_count,
u64 abort_count, u64 cycles_count,
u64 iter_count, u64 iter_cycles,
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
u64 from_count,
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
struct branch_type_stat *brtype_stat)
{
int printed;
if (branch_count == 0)
return scnprintf(bf, bfsize, " (calltrace)");
if (brtype_stat->branch_to) {
printed = branch_to_str(bf, bfsize, branch_count,
predicted_count, abort_count, brtype_stat);
} else {
printed = branch_from_str(bf, bfsize, branch_count,
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
cycles_count, iter_count, iter_cycles,
from_count);
perf report: Tag branch type/flag on "to" and tag cycles on "from" Current --branch-history LBR annotation displays confused data. For example, each cycles report is duplicated on both "from" and "to" entries. For example: perf report --branch-history --no-children --stdio --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7% cycles:1) main div.c:44 (predicted:49.7% cycles:1) main div.c:42 (RET CROSS_2M cycles:2) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M cycles:1) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M cycles:1) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M cycles:1) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M cycles:9) The cycles should be tagged only on the "from". It's for the code block that ends with "from", not for "to". Another issue is the "predicted:49.7%" is duplicated too (tag on both "from" and "to"). This patch tags the branch type/flag on "to" and tag the cycles on "from". For example: --2.32%--main div.c:39 (COND_BWD CROSS_2M predicted:49.7%) main div.c:44 (cycles:1) main div.c:42 (RET CROSS_2M) compute_flag div.c:28 (cycles:2) compute_flag div.c:27 (RET CROSS_2M) rand rand.c:28 (cycles:1) rand rand.c:28 (RET CROSS_2M) __random random.c:298 (cycles:1) __random random.c:297 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (COND_BWD CROSS_2M) __random random.c:295 (cycles:1) __random random.c:295 (RET CROSS_2M) | --2.23%--__random_r random_r.c:392 (cycles:9) In this example, The "main div.c:39 (COND_BWD CROSS_2M predicted:49.7%)" is "to" of branch and "main div.c:44 (cycles:1)" is "from" of branch. It should be easier for understanding than before. Signed-off-by: Yao Jin <yao.jin@linux.intel.com> Reviewed-by: Andi Kleen <ak@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1500894547-18411-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2017-07-24 19:09:07 +08:00
}
if (!printed)
bf[0] = 0;
return printed;
}
static int callchain_counts_printf(FILE *fp, char *bf, int bfsize,
u64 branch_count, u64 predicted_count,
u64 abort_count, u64 cycles_count,
u64 iter_count, u64 iter_cycles,
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
u64 from_count,
struct branch_type_stat *brtype_stat)
{
char str[256];
counts_str_build(str, sizeof(str), branch_count,
predicted_count, abort_count, cycles_count,
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
iter_count, iter_cycles, from_count, brtype_stat);
if (fp)
return fprintf(fp, "%s", str);
return scnprintf(bf, bfsize, "%s", str);
}
int callchain_list_counts__printf_value(struct callchain_list *clist,
FILE *fp, char *bf, int bfsize)
{
u64 branch_count, predicted_count;
u64 abort_count, cycles_count;
u64 iter_count, iter_cycles;
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
u64 from_count;
branch_count = clist->branch_count;
predicted_count = clist->predicted_count;
abort_count = clist->abort_count;
cycles_count = clist->cycles_count;
iter_count = clist->iter_count;
iter_cycles = clist->iter_cycles;
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
from_count = clist->from_count;
return callchain_counts_printf(fp, bf, bfsize, branch_count,
predicted_count, abort_count,
cycles_count, iter_count, iter_cycles,
perf report: Fix wrong iteration count in --branch-history By calculating the removed loops, we can get the iteration count. But the iteration count could be reported incorrectly, reporting impossibly high counts. That's because previous code uses the number of removed LBR entries for the iteration count. That's not good. Fix this by increasing the iteration count when a loop is detected. When matching the chain, the iteration count would be added up, finally we need to compute the average value when printing out. For example, $ perf report --branch-history --stdio --no-children Before: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:2968 avg_cycles:3) | f1 +0 | main +22 (cycles:1 iter:2968 avg_cycles:3) | main +17 | main +38 (cycles:1 iter:2968 avg_cycles:3) 2968 is an impossible high iteration count and avg_cycles is too small. After: ---f2 +0 | |--33.62%--f1 +9 (cycles:1) | f1 +0 | main +22 (cycles:1) | main +17 | main +38 (cycles:1) | main +27 | f1 +26 (cycles:1) | f1 +24 | f2 +27 (cycles:7) | f2 +0 | f1 +19 (cycles:1) | f1 +14 | f2 +27 (cycles:11) | f2 +0 | f1 +9 (cycles:1 iter:1 avg_cycles:23) | f1 +0 | main +22 (cycles:1 iter:1 avg_cycles:23) | main +17 | main +38 (cycles:1 iter:1 avg_cycles:23) avg_cycles:23 is the average cycles of this iteration. Fixes: c4ee06251d42 ("perf report: Calculate the average cycles of iterations") Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1546582230-17507-1-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-01-04 14:10:30 +08:00
from_count, &clist->brtype_stat);
}
static void free_callchain_node(struct callchain_node *node)
{
struct callchain_list *list, *tmp;
struct callchain_node *child;
struct rb_node *n;
perf hists browser: Support flat callchains The flat callchain mode is to print all chains in a single, simple hierarchy so make it easy to see. Currently perf report --tui doesn't show flat callchains properly. With flat callchains, only leaf nodes are added to the final rbtree so it should show entries in parent nodes. To do that, add parent_val list to struct callchain_node and show them along with the (normal) val list. For example, consider following callchains with '-g graph'. $ perf report -g graph - 39.93% swapper [kernel.vmlinux] [k] intel_idle intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle - cpu_startup_entry 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel Before: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel After: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle - 28.63% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary - 11.30% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_kernel x86_64_start_reservations x86_64_start_kernel Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-8-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:43 +08:00
list_for_each_entry_safe(list, tmp, &node->parent_val, list) {
list_del_init(&list->list);
map__zput(list->ms.map);
perf hists browser: Support flat callchains The flat callchain mode is to print all chains in a single, simple hierarchy so make it easy to see. Currently perf report --tui doesn't show flat callchains properly. With flat callchains, only leaf nodes are added to the final rbtree so it should show entries in parent nodes. To do that, add parent_val list to struct callchain_node and show them along with the (normal) val list. For example, consider following callchains with '-g graph'. $ perf report -g graph - 39.93% swapper [kernel.vmlinux] [k] intel_idle intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle - cpu_startup_entry 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel Before: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel After: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle - 28.63% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary - 11.30% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_kernel x86_64_start_reservations x86_64_start_kernel Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-8-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:43 +08:00
free(list);
}
list_for_each_entry_safe(list, tmp, &node->val, list) {
list_del_init(&list->list);
map__zput(list->ms.map);
free(list);
}
n = rb_first(&node->rb_root_in);
while (n) {
child = container_of(n, struct callchain_node, rb_node_in);
n = rb_next(n);
rb_erase(&child->rb_node_in, &node->rb_root_in);
free_callchain_node(child);
free(child);
}
}
void free_callchain(struct callchain_root *root)
{
if (!symbol_conf.use_callchain)
return;
free_callchain_node(&root->node);
}
perf hists browser: Support flat callchains The flat callchain mode is to print all chains in a single, simple hierarchy so make it easy to see. Currently perf report --tui doesn't show flat callchains properly. With flat callchains, only leaf nodes are added to the final rbtree so it should show entries in parent nodes. To do that, add parent_val list to struct callchain_node and show them along with the (normal) val list. For example, consider following callchains with '-g graph'. $ perf report -g graph - 39.93% swapper [kernel.vmlinux] [k] intel_idle intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle - cpu_startup_entry 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel Before: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel After: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle - 28.63% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary - 11.30% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_kernel x86_64_start_reservations x86_64_start_kernel Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-8-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:43 +08:00
static u64 decay_callchain_node(struct callchain_node *node)
{
struct callchain_node *child;
struct rb_node *n;
u64 child_hits = 0;
n = rb_first(&node->rb_root_in);
while (n) {
child = container_of(n, struct callchain_node, rb_node_in);
child_hits += decay_callchain_node(child);
n = rb_next(n);
}
node->hit = (node->hit * 7) / 8;
node->children_hit = child_hits;
return node->hit;
}
void decay_callchain(struct callchain_root *root)
{
if (!symbol_conf.use_callchain)
return;
decay_callchain_node(&root->node);
}
perf hists browser: Support flat callchains The flat callchain mode is to print all chains in a single, simple hierarchy so make it easy to see. Currently perf report --tui doesn't show flat callchains properly. With flat callchains, only leaf nodes are added to the final rbtree so it should show entries in parent nodes. To do that, add parent_val list to struct callchain_node and show them along with the (normal) val list. For example, consider following callchains with '-g graph'. $ perf report -g graph - 39.93% swapper [kernel.vmlinux] [k] intel_idle intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle - cpu_startup_entry 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel Before: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel After: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle - 28.63% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary - 11.30% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_kernel x86_64_start_reservations x86_64_start_kernel Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-8-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:43 +08:00
int callchain_node__make_parent_list(struct callchain_node *node)
{
struct callchain_node *parent = node->parent;
struct callchain_list *chain, *new;
LIST_HEAD(head);
while (parent) {
list_for_each_entry_reverse(chain, &parent->val, list) {
new = malloc(sizeof(*new));
if (new == NULL)
goto out;
*new = *chain;
new->has_children = false;
map__get(new->ms.map);
perf hists browser: Support flat callchains The flat callchain mode is to print all chains in a single, simple hierarchy so make it easy to see. Currently perf report --tui doesn't show flat callchains properly. With flat callchains, only leaf nodes are added to the final rbtree so it should show entries in parent nodes. To do that, add parent_val list to struct callchain_node and show them along with the (normal) val list. For example, consider following callchains with '-g graph'. $ perf report -g graph - 39.93% swapper [kernel.vmlinux] [k] intel_idle intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle - cpu_startup_entry 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel Before: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel After: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle - 28.63% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary - 11.30% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_kernel x86_64_start_reservations x86_64_start_kernel Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-8-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:43 +08:00
list_add_tail(&new->list, &head);
}
parent = parent->parent;
}
list_for_each_entry_safe_reverse(chain, new, &head, list)
list_move_tail(&chain->list, &node->parent_val);
if (!list_empty(&node->parent_val)) {
chain = list_first_entry(&node->parent_val, struct callchain_list, list);
chain->has_children = rb_prev(&node->rb_node) || rb_next(&node->rb_node);
chain = list_first_entry(&node->val, struct callchain_list, list);
chain->has_children = false;
}
return 0;
out:
list_for_each_entry_safe(chain, new, &head, list) {
list_del_init(&chain->list);
map__zput(chain->ms.map);
perf hists browser: Support flat callchains The flat callchain mode is to print all chains in a single, simple hierarchy so make it easy to see. Currently perf report --tui doesn't show flat callchains properly. With flat callchains, only leaf nodes are added to the final rbtree so it should show entries in parent nodes. To do that, add parent_val list to struct callchain_node and show them along with the (normal) val list. For example, consider following callchains with '-g graph'. $ perf report -g graph - 39.93% swapper [kernel.vmlinux] [k] intel_idle intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle - cpu_startup_entry 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel Before: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle 28.63% start_secondary - 11.30% rest_init start_kernel x86_64_start_reservations x86_64_start_kernel After: $ perf report -g flat - 39.93% swapper [kernel.vmlinux] [k] intel_idle - 28.63% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_secondary - 11.30% intel_idle cpuidle_enter_state cpuidle_enter call_cpuidle cpu_startup_entry start_kernel x86_64_start_reservations x86_64_start_kernel Signed-off-by: Namhyung Kim <namhyung@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kan Liang <kan.liang@intel.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1447047946-1691-8-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-11-09 13:45:43 +08:00
free(chain);
}
return -ENOMEM;
}
int callchain_cursor__copy(struct callchain_cursor *dst,
struct callchain_cursor *src)
{
int rc = 0;
callchain_cursor_reset(dst);
callchain_cursor_commit(src);
while (true) {
struct callchain_cursor_node *node;
node = callchain_cursor_current(src);
if (node == NULL)
break;
rc = callchain_cursor_append(dst, node->ip, &node->ms,
node->branch, &node->branch_flags,
node->nr_loop_iter,
node->iter_cycles,
node->branch_from, node->srcline);
if (rc)
break;
callchain_cursor_advance(src);
}
return rc;
}
/*
* Initialize a cursor before adding entries inside, but keep
* the previously allocated entries as a cache.
*/
void callchain_cursor_reset(struct callchain_cursor *cursor)
{
struct callchain_cursor_node *node;
cursor->nr = 0;
cursor->last = &cursor->first;
for (node = cursor->first; node != NULL; node = node->next)
map__zput(node->ms.map);
}
void callchain_param_setup(u64 sample_type, const char *arch)
{
if (symbol_conf.use_callchain || symbol_conf.cumulate_callchain) {
if ((sample_type & PERF_SAMPLE_REGS_USER) &&
(sample_type & PERF_SAMPLE_STACK_USER)) {
callchain_param.record_mode = CALLCHAIN_DWARF;
dwarf_callchain_users = true;
} else if (sample_type & PERF_SAMPLE_BRANCH_STACK)
callchain_param.record_mode = CALLCHAIN_LBR;
else
callchain_param.record_mode = CALLCHAIN_FP;
}
/*
* It's necessary to use libunwind to reliably determine the caller of
* a leaf function on aarch64, as otherwise we cannot know whether to
* start from the LR or FP.
*
* Always starting from the LR can result in duplicate or entirely
* erroneous entries. Always skipping the LR and starting from the FP
* can result in missing entries.
*/
if (callchain_param.record_mode == CALLCHAIN_FP && !strcmp(arch, "arm64"))
dwarf_callchain_users = true;
}
static bool chain_match(struct callchain_list *base_chain,
struct callchain_list *pair_chain)
{
enum match_result match;
match = match_chain_strings(base_chain->srcline,
pair_chain->srcline);
if (match != MATCH_ERROR)
return match == MATCH_EQ;
match = match_chain_dso_addresses(base_chain->ms.map,
base_chain->ip,
pair_chain->ms.map,
pair_chain->ip);
return match == MATCH_EQ;
}
bool callchain_cnode_matched(struct callchain_node *base_cnode,
struct callchain_node *pair_cnode)
{
struct callchain_list *base_chain, *pair_chain;
bool match = false;
pair_chain = list_first_entry(&pair_cnode->val,
struct callchain_list,
list);
list_for_each_entry(base_chain, &base_cnode->val, list) {
if (&pair_chain->list == &pair_cnode->val)
return false;
if (!base_chain->srcline || !pair_chain->srcline) {
pair_chain = list_next_entry(pair_chain, list);
continue;
}
match = chain_match(base_chain, pair_chain);
if (!match)
return false;
pair_chain = list_next_entry(pair_chain, list);
}
/*
* Say chain1 is ABC, chain2 is ABCD, we consider they are
* not fully matched.
*/
if (pair_chain && (&pair_chain->list != &pair_cnode->val))
return false;
return match;
}
static u64 count_callchain_hits(struct hist_entry *he)
{
struct rb_root *root = &he->sorted_chain;
struct rb_node *rb_node = rb_first(root);
struct callchain_node *node;
u64 chain_hits = 0;
while (rb_node) {
node = rb_entry(rb_node, struct callchain_node, rb_node);
chain_hits += node->hit;
rb_node = rb_next(rb_node);
}
return chain_hits;
}
u64 callchain_total_hits(struct hists *hists)
{
struct rb_node *next = rb_first_cached(&hists->entries);
u64 chain_hits = 0;
while (next) {
struct hist_entry *he = rb_entry(next, struct hist_entry,
rb_node);
chain_hits += count_callchain_hits(he);
next = rb_next(&he->rb_node);
}
return chain_hits;
}
s64 callchain_avg_cycles(struct callchain_node *cnode)
{
struct callchain_list *chain;
s64 cycles = 0;
list_for_each_entry(chain, &cnode->val, list) {
if (chain->srcline && chain->branch_count)
cycles += chain->cycles_count / chain->branch_count;
}
return cycles;
}