By using a mutex just for inserting and rotating two hist_entry rb
trees, so that when sorting we can get the last batch of entries created
from the ring buffer, merge it with whatever we have processed so far
and show the output while new entries are being added.
The 'report' tool continues, for now, to do it without threading, but
will use this in the future to allow visualization of results in long
perf.data sessions while the entries are being processed.
The new 'top' tool will be the first user.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-9b05atsn0q6m7fqgrug8fk2i@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
These are probably some old leftovers.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Sam Liao <phyomh@gmail.com>
These don't need to be globally visible.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Sam Liao <phyomh@gmail.com>
In order to implement callchains collapsing, we need to keep
track of the maximum depth in a histogram tree of callchains.
This way we'll avoid allocating an arbitrary temporary buffer
size on callchain merge time.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Christoph Hellwig <hch@infradead.org>
The stock newt checkbox tree widget we were using was not really
suitable for hist entry + callchain browsing.
The problems with it were manifold:
- We needed to traverse the whole hist_entry rb_tree to add each entry +
callchains beforehand.
- No control over the colors used for each row
So a new tree widget, based mostly on slang, was written.
It extends the ui_browser class already used for annotate to allow the
user to fold/unfold branches in the callchains tree, using extra fields
in the symbol_map class that is embedded in hist_entry and
callchain_node instances to store the folding state and when changing
this state calculates the number of rows that are produced when showing
a particular hist_entry instance.
This greatly speeds up browsing as we don't have to upfront touch all
the entries and only calculate callchain related operations when some
callchain branch is actually unfolded.
The memory footprint is also reduced as the data structure is not
duplicated, just some extra fields for controling callchain state and to
simplify the process of seeking thru entries (nr_rows, row_offset) were
added.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
They were globals, and since we support multiple hists and sessions
at the same time, it doesn't make sense to calculate those values
considereing all symbols in all sessions.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
In a shared multi-core environment, users want to analyze why their
program was slow. In particular, if the code ran slower only on certain
CPUs due to interference from other programs or kernel threads, the user
should be able to notice that.
Sample usage:
perf record -f -a -- sleep 3
perf report --sort cpu,comm
Workload:
program is running on 16 CPUs
Experiencing interference from an antagonist only on 4 CPUs.
Samples: 106218177676 cycles
Overhead CPU Command
........ ... ...............
6.25% 2 program
6.24% 6 program
6.24% 11 program
6.24% 5 program
6.24% 9 program
6.24% 10 program
6.23% 15 program
6.23% 7 program
6.23% 3 program
6.23% 14 program
6.22% 1 program
6.20% 13 program
3.17% 12 program
3.15% 8 program
3.14% 0 program
3.13% 4 program
3.11% 4 antagonist
3.11% 0 antagonist
3.10% 8 antagonist
3.07% 12 antagonist
Cc: David S. Miller <davem@davemloft.net>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <20100505181612.GA5091@sharma-home.net>
Signed-off-by: Arun Sharma <aruns@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
OPT_SET_INT was renamed to OPT_SET_UINT since the only use in these
tools is to set something that has an enum type, that is builtin
compatible with unsigned int.
Several string constifications were done to make OPT_STRING require a
const char * type.
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Number of samples is meaningless after we switched to auto-freq, so
report the number of events, i.e. not the sum of the different periods,
but the number PERF_RECORD_SAMPLE emitted by the kernel.
While doing this I noticed that naming "count" to the sum of all the
event periods can be confusing, so rename it to .period, just like in
struct sample.data, so that we become more consistent.
This helps with the next step, that was to record in struct hist_entry
the number of sample events for each instance, we need that because we
use it to generate the number of events when applying filters to the
tree of hist entries like it is being done in the TUI report browser.
Suggested-by: Ingo Molnar <mingo@elte.hu>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Now we don't anymore use popen to run 'perf annotate' for the selected
symbol, instead we collect per address samplings when processing samples
in 'perf report' if we're using the newt browser, then we use this data
directly to do annotation.
Done this way we can actually traverse the objdump_line objects
directly, matching the addresses to the collected samples and colouring
them appropriately using lower level slang routines.
The new ui_browser class will be reused for the main, callchain aware,
histogram browser, when it will be made generic and don't assume that
the objects are always instances of the objdump_line class maintained
using list_heads.
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
struct sort_entry has a callback named snprintf that turns an
entry into a string result.
But there are glibc versions that implement snprintf through a
macro. The following expression is then going to get the snprintf
call preprocessed:
ent->snprintf(...)
to finally end up in a build error:
util/hist.c: Dans la fonction «hist_entry__snprintf» :
util/hist.c:539: erreur: «struct sort_entry» has no member named «__builtin___snprintf_chk»
To fix this, prepend struct sort_entry callbacks with an "se_"
prefix.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Now one can press the right arrow key and in addition to being able to
filter by DSO, filter out by thread too, or a combination of both
filters.
With this one can start collecting events for the whole system, then
focus on a subset of the collected data quickly.
Cc: Avi Kivity <avi@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Clicking on -> will bring as one of the popup menu options a "Zoom into
CURRENT DSO", i.e. CURRENT will be replaced by the name of the DSO in
the current line.
Choosing this option will filter out all samples that didn't took place
in a symbol in this DSO.
After that the option reverts to "Zoom out of CURRENT DSO", to allow
going back to the more compreensive view, not filtered by DSO.
Future similar operations will include zooming into a particular thread,
COMM, CPU, "last minute", "last N usecs", etc.
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The struct callchain_node size is 120 bytes, that are never used when
there are no callchains or '-g none' is specified, so conditionally
allocate it, reducing sizeof(struct hist_entry) from 210 bytes to only
96, greatly speeding the non-callchain processing.
LKML-Reference: <new-submission>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Then hist_entry__fprintf will just us the newly introduced
hist_entry__snprintf, add the newline and fprintf it to the supplied
FILE descriptor.
This allows us to remove the use_browser checking in the color_printf
routines, that now got color_snprintf variants too.
The newt TUI browser (and other GUIs that may come in the future) don't
have to worry about stdio specific stuff in the strings they get from
the se->snprintf routines and instead use whatever means to do the
equivalent.
Also the newt TUI browser don't have to use the fmemopen() hack, instead
it can use the se->snprintf routines directly. For now tho use the
hist_entry__snprintf routine to reduce the patch size.
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
That will be in both struct hist_entry and struct
callchain_list, so that the TUI can store a pointer to the pair
(map, symbol) in the trees where hist_entries and
callchain_lists are present, to allow precise annotation instead
of looking for the first symbol with the selected name.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1269459619-982-4-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
That means that almost everything you can do with 'perf report'
can be done with 'perf diff', for instance:
$ perf record -f find / > /dev/null
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.062 MB perf.data (~2699
samples) ] $ perf record -f find / > /dev/null
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.062 MB perf.data (~2687
samples) ] perf diff | head -8
9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal
2.91% -1.00% find [kernel] [k] __kmalloc
2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent
1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock
2.44% find [kernel] [k] half_md4_transform
$
So if you want to zoom into libc:
$ perf diff --dsos libc-2.10.1.so | head -8
37.34% find [.] _IO_vfprintf_internal
10.34% find [.] __GI_memmove
8.25% +2.00% find [.] _int_malloc
5.07% -1.00% find [.] __GI_mempcpy
7.62% +2.00% find [.] _int_free
$
And if there were multiple commands using libc, it is also
possible to aggregate them all by using --sort symbol:
$ perf diff --dsos libc-2.10.1.so --sort symbol | head -8
37.34% [.] _IO_vfprintf_internal
10.34% [.] __GI_memmove
8.25% +2.00% [.] _int_malloc
5.07% -1.00% [.] __GI_mempcpy
7.62% +2.00% [.] _int_free
$
The displacement column now is off by default, to use it:
perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8
37.34% [.] _IO_vfprintf_internal
10.34% [.] __GI_memmove
8.25% +2.00% [.] _int_malloc
5.07% -1.00% +2 [.] __GI_mempcpy
7.62% +2.00% -1 [.] _int_free
$
Using -t/--field-separator can be used for scripting:
$ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8
37.34, , ,[.] _IO_vfprintf_internal
10.34, , ,[.] __GI_memmove
8.25,+2.00%, ,[.] _int_malloc
5.07,-1.00%, +2,[.] __GI_mempcpy
7.62,+2.00%, -1,[.] _int_free
6.99,+1.00%, -1,[.] _IO_new_file_xsputn
1.89,-2.00%, +4,[.] __readdir64
$
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
And it is also needed by 'perf diff'.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1260828571-3613-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Currently, the callchains are displayed using a constant left
margin. So depending on the current sort dimension
configuration, callchains may appear to be well attached to the
first sort dimension column field which is mostly the case,
except when the first dimension of sorting is done by comm,
because these are right aligned.
This patch binds the callchain to the first letter in the first
column, whatever type of column it is (dso, comm, symbol).
Before:
0.80% perf [k] __lock_acquire
__lock_acquire
lock_acquire
|
|--58.33%-- _spin_lock
| |
| |--28.57%-- inotify_should_send_event
| | fsnotify
| | __fsnotify_parent
After:
0.80% perf [k] __lock_acquire
__lock_acquire
lock_acquire
|
|--58.33%-- _spin_lock
| |
| |--28.57%-- inotify_should_send_event
| | fsnotify
| | __fsnotify_parent
Also, for clarity, we don't put anymore the callchain as is but:
- If we have a top level ancestor in the callchain, start it
with a first ascii hook.
Before:
0.80% perf [kernel] [k] __lock_acquire
__lock_acquire
lock_acquire
|
|--58.33%-- _spin_lock
| |
| |--28.57%-- inotify_should_send_event
| | fsnotify
[..] [..]
After:
0.80% perf [kernel] [k] __lock_acquire
|
--- __lock_acquire
lock_acquire
|
|--58.33%-- _spin_lock
| |
| |--28.57%-- inotify_should_send_event
| | fsnotify
[..] [..]
- Otherwise, if we have several top level ancestors, then
display these like we did before:
1.69% Xorg
|
|--21.21%-- vread_hpet
| 0x7fffd85b46fc
| 0x7fffd85b494d
| 0x7f4fafb4e54d
|
|--15.15%-- exaOffscreenAlloc
|
|--9.09%-- I830WaitLpRing
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Anton Blanchard <anton@samba.org>
LKML-Reference: <1256246604-17156-2-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
While recursively printing the branches of each callchains, we
forget to display the root. It is never printed.
Say we have:
symbol
f1
f2
|
-------- f3
| f4
|
---------f5
f6
Actually we never see that, instead it displays:
symbol
|
--------- f3
| f4
|
--------- f5
f6
However f1 is always the same than "symbol" and if we are
sorting by symbols first then "symbol", f1 and f2 will be well
aligned like in the above example, so displaying f1 looks
redundant here.
But if we are sorting by something else first (dso, comm,
etc...), displaying f1 doesn't look redundant but rather
necessary because the symbol is not well aligned anymore with
its callchain:
comm dso symbol
f1
f2
|
--------- [...]
And we want the callchain to be obvious.
So we fix the bug by printing the root branch, but we also
filter its first entry if we are sorting by symbols first.
Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1256246604-17156-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Representing modules as struct map entries, backed by a DSO, etc,
using /proc/modules to find where the module is loaded.
DSOs now can have a short and long name, so that in verbose mode we
can show exactly which .ko or vmlinux image was used.
As kernel modules now are a DSO separate from the kernel, we can
ask for just the hits for a particular set of kernel modules, just
like we can do with shared libraries:
[root@doppio linux-2.6-tip]# perf report -n --vmlinux
/home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15
84.58% 13266 Xorg [k] drm_clflush_pages
4.02% 630 Xorg [k] trace_kmalloc.clone.0
3.95% 619 Xorg [k] drm_ioctl
2.07% 324 Xorg [k] drm_addbufs
1.68% 263 Xorg [k] drm_gem_close_ioctl
0.77% 120 Xorg [k] drm_setmaster_ioctl
0.70% 110 Xorg [k] drm_lastclose
0.68% 106 Xorg [k] drm_open
0.54% 85 Xorg [k] drm_mm_search_free
[root@doppio linux-2.6-tip]#
Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko
would have the same effect. Allowing specifying just 'drm.ko' is left
for another patch.
Processing kallsyms so that per kernel module struct map are
instantiated was also left for another patch. That will allow
removing the module name from each of its symbols.
struct symbol was reduced by removing the ->module backpointer and
moving it (well now the map) to struct symbol_entry in perf top,
that is its only user right now.
The total linecount went down by ~500 lines.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Avi Kivity <avi@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Create util/sort.[ch] and move common functionality for
builtin-report.c and builtin-annotate.c there, and make use of it.
Signed-off-by: John Kacur <jkacur@redhat.com>
LKML-Reference: <alpine.LFD.2.00.0909241758390.11383@localhost.localdomain>
Signed-off-by: Ingo Molnar <mingo@elte.hu>