This commit adds support for a new modifier "D", which requests that the
event, or group of events, be pinned to the PMU.
The "p" modifier is already taken for precise, and "P" may be used in
future to mean "fully precise".
So we use "D", which stands for pinneD - and looks like a padlock, or if
you're using the ":D" syntax perf smiles at you.
This is an oft-requested feature from our HW folks, who want to be able
to run a large number of events, but also want 100% accurate results for
instructions per cycle.
Comparison of results with and without pinning:
$ perf stat -e '{cycles,instructions}:D' -e cycles,instructions,...
79,590,480,683 cycles # 0.000 GHz
166,123,716,524 instructions # 2.09 insns per cycle
# 0.11 stalled cycles per insn
79,352,134,463 cycles # 0.000 GHz [11.11%]
165,178,301,818 instructions # 2.08 insns per cycle
# 0.11 stalled cycles per insn [11.13%]
As you can see although perf does a very good job of scaling the values
in the non-pinned case, there is some small discrepancy.
The patch is fairly straight forward, the one detail is that we need to
make sure we only request pinning for the group leader when we have a
group.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Tested-by: Jiri Olsa <jolsa@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1375795686-4226-1-git-send-email-michael@ellerman.id.au
[ Use perf_evsel__is_group_leader instead of open coded equivalent, as
suggested by Jiri Olsa ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When measuring workloads the startup phase -- doing page faults, dynamic
linking, opening files -- is often very different from the rest of the
workload. Especially with smaller kernels and using counter
multiplexing this can give significant measurement errors.
Multiplexing assumes that the workload is mostly the same over longer
periods. But at startup there is typically some spike of activity which
is relatively short. If many groups are multiplexing the one group
seeing the spike, and which is then scaled up over the time to run all
groups, may see a significant error.
Also in general it's often not useful to measure the startup, because it
is so different from the rest.
One way around this is to use interval mode and discard the first
sample, but this can be awkward because interval mode doesn't support
intervals of less than 100ms, and also a useful interval is not
necessarily the same as a useful startup delay.
This patch adds a new --initial-delay / -D option to skip measuring for
the startup phase. The time can be specified in ms
Here's a simple example:
perf stat -e page-faults bash -c 'for i in $(seq 100000) ; do true ; done'
...
3,721 page-faults
...
If we just wait 20 ms the number of page faults is 1/3 less:
perf stat -D 20 -e page-faults bash -c 'for i in $(seq 100000) ; do true ; done'
...
2,823 page-faults
...
So we filtered out most of the startup noise from bash.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1375490473-1503-4-git-send-email-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Adding 'S' event/group modifier to specify that the event value/s are
read by PERF_SAMPLE_READ sample type processing, instead of the period
value offered by lower layers.
There's additional behaviour change for 'S' modifier being specified on
event group:
Currently all the events within a group makes samples. If user now
specifies 'S' within group modifier, only the leader will trigger
samples. The rest of events in the group will have sampling disabled.
And same as for single events, values of all events within the group
(including leader) are read by PERF_SAMPLE_READ sample type processing.
Following example will create event group with cycles and cache-misses
events, setting the cycles as group leader and the only event to
actually sample. Both cycles and cache-misses event period values are
read by PERF_SAMPLE_READ sample type processing with PERF_FORMAT_GROUP
read format.
Example:
$ perf record -e '{cycles,cache-misses}:S' ls
...
$ perf report --group --show-total-period --stdio
...
# Samples: 36 of event 'anon group { cycles, cache-misses }'
# Event count (approx.): 12585593
#
# Overhead Period Command Shared Object Symbol
# .............. .............. ....... ................. ..........................
#
19.92% 1.20% 2505936 31 ls [kernel.kallsyms] [k] mark_held_locks
13.74% 0.47% 1729327 12 ls [kernel.kallsyms] [k] sched_clock_local
13.64% 23.72% 1716147 612 ls ld-2.14.90.so [.] check_match.10805
13.12% 23.22% 1650778 599 ls libc-2.14.90.so [.] _nl_intern_locale_data
11.24% 29.19% 1414554 753 ls [kernel.kallsyms] [k] sched_clock_cpu
8.50% 0.35% 1070150 9 ls [kernel.kallsyms] [k] check_chain_key
...
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/n/tip-iyoinu3axi11mymwnh2b7fxj@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
With programs with very large functions it can be useful to distinguish
the callgraph nodes on more than just function names. So for example if
you have multiple calls to the same function, it ends up being separate
nodes in the chain.
This patch adds a new key field to the callgraph options, that allows
comparing nodes on functions (as today, default) and addresses.
Longer term it would be nice to also handle src lines, but that would
need more changes and address is a reasonable proxy for it today.
I right now reference the global params, as there was no simple way to
register a params pointer.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-0uskktybf0e7wrnoi5e9b9it@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
For example, in an application with an expensive function implemented
with deeply nested recursive calls, the default call-graph presentation
is dominated by the different callchains within that function. By
ignoring these callees, we can collect the callchains leading into the
function and compactly identify what to blame for expensive calls.
For example, in this report the callers of garbage_collect() are
scattered across the tree:
$ perf report -d ruby 2>- | grep -m10 ^[^#]*[a-z]
22.03% ruby [.] gc_mark
--- gc_mark
|--59.40%-- mark_keyvalue
| st_foreach
| gc_mark_children
| |--99.75%-- rb_gc_mark
| | rb_vm_mark
| | gc_mark_children
| | gc_marks
| | |--99.00%-- garbage_collect
If we ignore the callees of garbage_collect(), its callers are coalesced:
$ perf report --ignore-callees garbage_collect -d ruby 2>- | grep -m10 ^[^#]*[a-z]
72.92% ruby [.] garbage_collect
--- garbage_collect
vm_xmalloc
|--47.08%-- ruby_xmalloc
| st_insert2
| rb_hash_aset
| |--98.45%-- features_index_add
| | rb_provide_feature
| | rb_require_safe
| | vm_call_method
Signed-off-by: Greg Price <price@mit.edu>
Tested-by: Jiri Olsa <jolsa@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20130623031720.GW22203@biohazard-cafe.mit.edu
Link: http://lkml.kernel.org/r/20130708115746.GO22203@biohazard-cafe.mit.edu
Cc: Fengguang Wu <fengguang.wu@intel.com>
[ remove spaces at beginning of line, reported by Fengguang Wu ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf record has a new option -W that enables weightened sampling.
Add sorting support in top/report for the average weight per sample and the
total weight sum. This allows to both compare relative cost per event
and the total cost over the measurement period.
Add the necessary glue to perf report, record and the library.
v2: Merge with new hist refactoring.
v3: Fix manpage. Remove value check.
Rename global_weight to weight and weight to local_weight.
v4: Readd sort keys to manpage
v5: Move weight to end
v6: Move weight to template
v7: Rename weight key.
Original patch from Andi modified by Stephane Eranian <eranian@google.com>
to include ONLY the weight supporting code and apply to pristine 3.8.0-rc4.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1359040242-8269-6-git-send-email-eranian@google.com
[ committer note: changed to cope with fc5871ed and the hists_link perf test entry ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This patch adds the --per-core option to perf stat.
This option is used to aggregate system-wide counts
on a per physical core basis. On processors with
hyperthreading, this means counts of all HT threads
running on a physical core are aggregated.
This mode is useful to find imblance between physical
cores running an uniform workload. Cores are identified
by socket: S0-C1, means physical core 1 on socket 0. Note
that cores are identified using their physical core id,
thus their numbering may not be continuous.
Per core aggregation can be combined with interval printing:
# perf stat -a --per-core -I 1000 -e cycles sleep 1000
# time core cpus counts events
1.000090030 S0-C0 1 4,765,747 cycles
1.000090030 S0-C1 1 5,580,647 cycles
1.000090030 S0-C2 1 221,181 cycles
1.000090030 S0-C3 1 266,092 cycles
Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1360846649-6411-4-git-send-email-eranian@google.com
[ committer note: Remove parts already applied on 86ee6e1 to keep bisectability ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This patch adds per-processor socket count aggregation for system-wide
mode measurements. This is a useful mode to detect imbalance between
sockets.
To enable this mode, use --aggr-socket in addition
to -a. (system-wide).
The output includes the socket number and the number of online
processors on that socket. This is useful to gauge the amount of
aggregation.
# ./perf stat -I 1000 -a --aggr-socket -e cycles sleep 2
# time socket cpus counts events
1.000097680 S0 4 5,788,785 cycles
2.000379943 S0 4 27,361,546 cycles
2.001167808 S0 4 818,275 cycles
Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1360161962-9675-3-git-send-email-eranian@google.com
[ committer note: Added missing man page entry based on above comments ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This patch adds a new printing mode for perf stat. It allows interval
printing. That means perf stat can now print event deltas at regular
time interval. This is useful to detect phases in programs.
The -I option enables interval printing. It expects an interval duration
in milliseconds. Minimum is 100ms. Once, activated perf stat prints
events deltas since last printout. All modes are supported.
$ perf stat -I 1000 -e cycles noploop 10
noploop for 10 seconds
# time counts events
1.000109853 2,388,560,546 cycles
2.000262846 2,393,332,358 cycles
3.000354131 2,393,176,537 cycles
4.000439503 2,393,203,790 cycles
5.000527075 2,393,167,675 cycles
6.000609052 2,393,203,670 cycles
7.000691082 2,393,175,678 cycles
The output format makes it easy to feed into a plotting program such as
gnuplot when the -I option is used in combination with the -x option:
$ perf stat -x, -I 1000 -e cycles noploop 10
noploop for 10 seconds
1.000084113,2378775498,cycles
2.000245798,2391056897,cycles
3.000354445,2392089414,cycles
4.000459115,2390936603,cycles
5.000565341,2392108173,cycles
Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1359460064-3060-3-git-send-email-eranian@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Sometimes a test is problematic for some reason and one wants to skip it,
for instance:
[root@sandy ~]# perf test
1: vmlinux symtab matches kallsyms : Ok
2: detect open syscall event : Ok
3: detect open syscall event on all cpus : Ok
4: read samples using the mmap interface : Ok
5: parse events tests : Warning: bad op token {
Warning: bad op token {
Warning: bad op token {
Warning: bad op token {
Warning: bad op token {
Warning: function is_writable_pte not defined
Segmentation fault (core dumped)
So now we can use -s/--skip while the problematic tests are being fixed,
allowing us to test all the other entries:
[root@sandy ~]# perf test -s 5
1: vmlinux symtab matches kallsyms : Ok
2: detect open syscall event : Ok
3: detect open syscall event on all cpus : Ok
4: read samples using the mmap interface : Ok
5: parse events tests : Skip (user override)
6: x86 rdpmc test : Ok
7: Validate PERF_RECORD_* events & perf_sample fields : Ok
8: Test perf pmu format parsing : Ok
9: Test dso data interface : Ok
10: roundtrip evsel->name check : Ok
11: Check parsing of sched tracepoints fields : Ok
12: Generate and check syscalls:sys_enter_open event fields: Ok
13: struct perf_event_attr setup : Ok
14: Test matching and linking mutliple hists : Ok
15: Try 'use perf' in python, checking link problems : Ok
[root@sandy ~]#
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-klzd8p57jzdryafqkmlppcb1@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:
. perf build-id cache now can show DSOs present in a perf.data file that are
not in the cache, to integrate with build-id servers being put in place by
organizations such as Fedora.
. perf buildid-list -i an-elf-file-instead-of-a-perf.data is back showing its
build-id.
. No need to do feature checks when doing a 'make tags'
. Fix some 'perf test' errors and make them use the tracepoint evsel constructor.
. perf top now shares more of the evsel config/creation routines with 'record',
paving the way for further integration like 'top' snapshots, etc.
. perf top now supports DWARF callchains.
. perf evlist decodes sample_type and read_format, helping diagnose problems.
. Fix mmap limitations on 32-bit, fix from David Miller.
. perf diff fixes from Jiri Olsa.
. Ignore ABS symbols when loading data maps, fix from Namhyung Kim
. Hists improvements from Namhyung Kim
. Don't check configuration on make clean, from Namhyung Kim
. Fix dso__fprintf() print statement, from Stephane Eranian.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull trivial branch from Jiri Kosina:
"Usual stuff -- comment/printk typo fixes, documentation updates, dead
code elimination."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial: (39 commits)
HOWTO: fix double words typo
x86 mtrr: fix comment typo in mtrr_bp_init
propagate name change to comments in kernel source
doc: Update the name of profiling based on sysfs
treewide: Fix typos in various drivers
treewide: Fix typos in various Kconfig
wireless: mwifiex: Fix typo in wireless/mwifiex driver
messages: i2o: Fix typo in messages/i2o
scripts/kernel-doc: check that non-void fcts describe their return value
Kernel-doc: Convention: Use a "Return" section to describe return values
radeon: Fix typo and copy/paste error in comments
doc: Remove unnecessary declarations from Documentation/accounting/getdelays.c
various: Fix spelling of "asynchronous" in comments.
Fix misspellings of "whether" in comments.
eisa: Fix spelling of "asynchronous".
various: Fix spelling of "registered" in comments.
doc: fix quite a few typos within Documentation
target: iscsi: fix comment typos in target/iscsi drivers
treewide: fix typo of "suport" in various comments and Kconfig
treewide: fix typo of "suppport" in various comments
...
In order to measure kernel builds, one has to do some pre/post cleanup
work in order to do the repeat build.
So provide --pre and --post command hooks to allow doing just that.
perf stat --repeat 10 --null --sync --pre 'make -s O=defconfig-build/clean' \
-- make -s -j64 O=defconfig-build/ bzImage
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Ingo Molnar <mingo@kernel.org>
Cc: Stephane Eranian <eranian@gmail.com>
Link: http://lkml.kernel.org/r/1350992414.13456.5.camel@twins
[ committer note: Added respective entries in Documentation/perf-stat.txt ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
You may want to know where and how long a task is sleeping. A callchain
may be found in sched_switch and a time slice in stat_iowait, so I add
handler in perf inject for merging this events.
My code saves sched_switch event for each process and when it meets
stat_iowait, it reports the sched_switch event, because this event
contains a correct callchain. By another words it replaces all
stat_iowait events on proper sched_switch events.
I use the next sequence of commands for testing:
perf record -e sched:sched_stat_sleep -e sched:sched_switch \
-e sched:sched_process_exit -g -o ~/perf.data.raw \
~/test-program
perf inject -v -s -i ~/perf.data.raw -o ~/perf.data
perf report --stdio -i ~/perf.data
100.00% foo [kernel.kallsyms] [k] __schedule
|
--- __schedule
schedule
|
|--79.75%-- schedule_hrtimeout_range_clock
| schedule_hrtimeout_range
| poll_schedule_timeout
| do_select
| core_sys_select
| sys_select
| system_call_fastpath
| __select
| __libc_start_main
|
--20.25%-- do_nanosleep
hrtimer_nanosleep
sys_nanosleep
system_call_fastpath
__GI___libc_nanosleep
__libc_start_main
And here is test-program.c:
#include<unistd.h>
#include<time.h>
#include<sys/select.h>
int main()
{
struct timespec ts1;
struct timeval tv1;
int i;
long s;
for (i = 0; i < 10; i++) {
ts1.tv_sec = 0;
ts1.tv_nsec = 10000000;
nanosleep(&ts1, NULL);
tv1.tv_sec = 0;
tv1.tv_usec = 40000;
select(0, NULL, NULL, NULL,&tv1);
}
return 1;
}
Signed-off-by: Andrew Vagin <avagin@openvz.org>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1344344165-369636-4-git-send-email-avagin@openvz.org
[ committer note: Made it use evsel->handler ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
