The "comm" column allows grouping events by the process command. It is
intended to group like programs, despite having different PIDs. But some
workloads may adjust their own command, so that a unique identifier
(e.g. a PID or some other numeric value) is part of the command name.
This destroys the utility of "comm", forcing perf to place each unique
process name into its own bucket, which can contribute to a
combinatorial explosion of memory use in perf report.
Create a less strict version of this column, which ignores digits when
comparing command names. Commands whose names are the same (ignoring
digits) are sorted into the same histogram buckets, and displayed with
the placeholder value "<N>" in the place of digits. For example,
hypothetical command names "kworker/1" "kworker/2" "kworker/3" would
sort into the same bucket and be represented as "kworker/<N>".
Committer testing:
$ perf report -s comm,comm_nodigit | grep -F "<N>"
0.01% CPU 6/TCG CPU <N>/TCG
0.01% kworker/53:2-mm kworker/<N>:<N>-mm
0.01% migration/24 migration/<N>
0.01% kworker/24:1-ev kworker/<N>:<N>-ev
0.01% llvmpipe-8 llvmpipe-<N>
Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Previously arch_support_sort_key and arch_perf_header_entry used a
weak symbol to compile as appropriate for x86 and powerpc. A
limitation to this is that the handling of a data file could vary in
cross-platform development. Change to using the perf_env of the
current session to determine the architecture kind and set the sort
key and header entries as appropriate.
Signed-off-by: Ian Rogers <irogers@google.com>
Link: https://lore.kernel.org/r/20250724163302.596743-23-irogers@google.com
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Sometimes we need to analyze the data in process level but current sort
keys only work on thread level. Let's add 'tgid' sort key for that as
'pid' is already taken for thread.
This will look mostly the same, but it only uses tgid instead of tid.
Here's an example of a process with two threads (thloop).
$ perf record -- perf test -w thloop
$ perf report --stdio -s tgid,pid -H
...
#
# Overhead Tgid:Command / Pid:Command
# ........... ..........................
#
100.00% 2018407:perf
50.34% 2018407:perf
49.66% 2018409:perf
Suggested-by: Stephane Eranian <eranian@google.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20250509210421.197245-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Sometimes it's useful to organize member fields in cache-line boundary.
The 'typecln' sort key is short for type-cacheline and to show samples
in each cacheline. The cacheline size is fixed to 64 for now, but it
can read the actual size once it saves the value from sysfs.
For example, you maybe want to which cacheline in a target is hot or
cold. The following shows members in the cfs_rq's first cache line.
$ perf report -s type,typecln,typeoff -H
...
- 2.67% struct cfs_rq
+ 1.23% struct cfs_rq: cache-line 2
+ 0.57% struct cfs_rq: cache-line 4
+ 0.46% struct cfs_rq: cache-line 6
- 0.41% struct cfs_rq: cache-line 0
0.39% struct cfs_rq +0x14 (h_nr_running)
0.02% struct cfs_rq +0x38 (tasks_timeline.rb_leftmost)
...
Committer testing:
# root@number:~# perf report -s type,typecln,typeoff -H --stdio
# Total Lost Samples: 0
#
# Samples: 5K of event 'cpu_atom/mem-loads,ldlat=5/P'
# Event count (approx.): 312251
#
# Overhead Data Type / Data Type Cacheline / Data Type Offset
# .............. ..................................................
#
<SNIP>
0.07% struct sigaction
0.05% struct sigaction: cache-line 1
0.02% struct sigaction +0x58 (sa_mask)
0.02% struct sigaction +0x78 (sa_mask)
0.03% struct sigaction: cache-line 0
0.02% struct sigaction +0x38 (sa_mask)
0.01% struct sigaction +0x8 (sa_mask)
<SNIP>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20240819233603.54941-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Some sort keys are meaningful only in a specific mode - like branch
stack and memory (data-src). Add the mode to skip unnecessary ones.
This will be used for 'perf mem report' later.
While at it, change the prefix for the -F/--fields option to remove
the duplicate part.
Before:
$ perf report -F
Error: switch `F' requires a value
Usage: perf report [<options>]
-F, --fields <key[,keys...]>
output field(s): overhead period sample overhead overhead_sys
overhead_us overhead_guest_sys overhead_guest_us overhead_children
sample period weight1 weight2 weight3 ins_lat retire_lat
...
After:
$ perf report -F
Error: switch `F' requires a value
Usage: perf report [<options>]
-F, --fields <key[,keys...]>
output field(s): overhead overhead_sys overhead_us
overhead_guest_sys overhead_guest_us overhead_children
sample period weight1 weight2 weight3 ins_lat retire_lat
...
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240731235505.710436-5-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add weight1, weight2 and weight3 fields to -F/--fields and their aliases
like 'ins_lat', 'p_stage_cyc' and 'retire_lat'. Note that they are in
the sort keys too but the difference is that output fields will sum up
the weight values and display the average.
In the sort key, users can see the distribution of weight value and I
think it's confusing we have local vs. global weight for the same weight.
For example, I experiment with mem-loads events to get the weights. On
my laptop, it seems only weight1 field is supported.
$ perf mem record -- perf test -w noploop
Let's look at the noploop function only. It has 7 samples.
$ perf script -F event,ip,sym,weight | grep noploop
# event weight ip sym
cpu/mem-loads,ldlat=30/P: 43 55b3c122bffc noploop
cpu/mem-loads,ldlat=30/P: 48 55b3c122bffc noploop
cpu/mem-loads,ldlat=30/P: 38 55b3c122bffc noploop <--- same weight
cpu/mem-loads,ldlat=30/P: 38 55b3c122bffc noploop <--- same weight
cpu/mem-loads,ldlat=30/P: 59 55b3c122bffc noploop
cpu/mem-loads,ldlat=30/P: 33 55b3c122bffc noploop
cpu/mem-loads,ldlat=30/P: 38 55b3c122bffc noploop <--- same weight
When you use the 'weight' sort key, it'd show entries with a separate
weight value separately. Also note that the first entry has 3 samples
with weight value 38, so they are displayed together and the weight
value is the sum of 3 samples (114 = 38 * 3).
$ perf report -n -s +weight | grep -e Weight -e noploop
# Overhead Samples Command Shared Object Symbol Weight
0.53% 3 perf perf [.] noploop 114
0.18% 1 perf perf [.] noploop 59
0.18% 1 perf perf [.] noploop 48
0.18% 1 perf perf [.] noploop 43
0.18% 1 perf perf [.] noploop 33
If you use 'local_weight' sort key, you can see the actual weight.
$ perf report -n -s +local_weight | grep -e Weight -e noploop
# Overhead Samples Command Shared Object Symbol Local Weight
0.53% 3 perf perf [.] noploop 38
0.18% 1 perf perf [.] noploop 59
0.18% 1 perf perf [.] noploop 48
0.18% 1 perf perf [.] noploop 43
0.18% 1 perf perf [.] noploop 33
But when you use the -F/--field option instead, you can see the average
weight for the while noploop function (as it won't group samples by
weight value and use the default 'comm,dso,sym' sort keys).
$ perf report -n -F +weight | grep -e Weight -e noploop
Warning:
--fields weight shows the average value unlike in the --sort key.
# Overhead Samples Weight1 Command Shared Object Symbol
1.23% 7 42.4 perf perf [.] noploop
The weight1 field shows the average value:
(38 * 3 + 59 + 48 + 43 + 33) / 7 = 42.4
Also it'd show the warning that 'weight' field has the average value.
Using 'weight1' can remove the warning.
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240411181718.2367948-3-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The typeoff sort key shows the data type name, offset and the name of
the field. This is useful to see which field in the struct is accessed
most frequently.
$ perf report -s type,typeoff --hierarchy --stdio
...
# Overhead Data Type / Data Type Offset
# ............ ............................
#
...
1.23% struct cfs_rq
0.19% struct cfs_rq +404 (throttle_count)
0.19% struct cfs_rq +0 (load.weight)
0.19% struct cfs_rq +336 (leaf_cfs_rq_list.next)
0.09% struct cfs_rq +272 (propagate)
0.09% struct cfs_rq +196 (removed.nr)
0.09% struct cfs_rq +80 (curr)
0.09% struct cfs_rq +544 (lt_b_children_throttled)
0.06% struct cfs_rq +320 (rq)
Committer testing:
Again with the perf.data from the previous csets:
# perf report --stdio -s type,typeoff
# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 4 of event 'cpu_atom/mem-loads,ldlat=30/P'
# Event count (approx.): 7
#
# Overhead Data Type Data Type Offset
# ........ ......... ................
#
42.86% struct list_head struct list_head +8 (prev)
42.86% (unknown) (unknown) +0 (no field)
14.29% char char +0 (no field)
#
# (Tip: To see callchains in a more compact form: perf report -g folded)
#
# perf report --stdio -s dso,type,typeoff
# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 4 of event 'cpu_atom/mem-loads,ldlat=30/P'
# Event count (approx.): 7
#
# Overhead Shared Object Data Type Data Type Offset
# ........ .................... ......... ................
#
42.86% [kernel.kallsyms] struct list_head struct list_head +8 (prev)
28.57% libc.so.6 (unknown) (unknown) +0 (no field)
14.29% [kernel.kallsyms] char char +0 (no field)
14.29% ld-linux-x86-64.so.2 (unknown) (unknown) +0 (no field)
#
# (Tip: If you have debuginfo enabled, try: perf report -s sym,srcline)
#
#
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: linux-toolchains@vger.kernel.org
Cc: linux-trace-devel@vger.kernel.org
Link: https://lore.kernel.org/r/20231213001323.718046-13-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add 'simd' sort field to visualize SIMD ops in 'perf report'.
Rows are labeled with the SIMD ISA, and the type of predicate (if any):
- [p] partial predicate
- [e] empty predicate (no elements in the vector being used)
Example with Arm SPE and SVE (Scalable Vector Extension):
#include <arm_sve.h>
double src[1025], dst[1025];
int main(void) {
svfloat64_t vc = svdup_f64(1);
for(;;)
for(int i = 0; i < 1025; i += svcntd())
{
svbool_t pg = svwhilelt_b64(i, 1025);
svfloat64_t vsrc = svld1(pg, &src[i]);
svfloat64_t vdst = svadd_x(pg, vsrc, vc);
svst1(pg, &dst[i], vdst);
}
return 0;
}
... compiled using "gcc-11 -march=armv8-a+sve -O3"
Profiling on a platform that implements FEAT_SVE and FEAT_SPEv1p1:
$ perf record -e arm_spe_0// -- ./a.out
$ perf report --itrace=i1i -s overhead,pid,simd,sym
Overhead Pid:Command Simd Symbol
........ ................ ....... ......................
53.76% 10758:program [.] main
46.14% 10758:program [.] SVE [.] main
0.09% 10758:program [p] SVE [.] main
The report shows 0.09% of the sampled SVE operations use partial
predicates due to src and dst arrays not being multiples of the vector
register lengths.
Signed-off-by: German Gomez <german.gomez@arm.com>
Acked-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Anshuman.Khandual@arm.com
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: https://lore.kernel.org/r/20230320151509.1137462-2-james.clark@arm.com
Signed-off-by: James Clark <james.clark@arm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Many platforms have feature of adjacent cachelines prefetch, when it is
enabled, for data in RAM of 2 cachelines (2N and 2N+1) granularity, if
one is fetched to cache, the other one could likely be fetched too,
which sort of extends the cacheline size to double, thus the false
sharing could happens in adjacent cachelines.
0Day has captured performance changed related with this [1], and some
commercial software explicitly makes its hot global variables 128 bytes
aligned (2 cache lines) to avoid this kind of extended false sharing.
So add an option "--double-cl" for 'perf c2c report' to show false
sharing in double cache line granularity, which acts just like the
cacheline size is doubled. There is no change to c2c record. The
hardware events of shared cacheline are still per cacheline, and this
option just changes the granularity of how events are grouped and
displayed.
In the 'perf c2c report' output below (will-it-scale's 'pagefault2' case
on old kernel):
----------------------------------------------------------------------
26 31 2 0 0 0 0xffff888103ec6000
----------------------------------------------------------------------
35.48% 50.00% 0.00% 0.00% 0.00% 0x10 0 1 0xffffffff8133148b 1153 66 971 3748 74 [k] get_mem_cgroup_from_mm
6.45% 0.00% 0.00% 0.00% 0.00% 0x10 0 1 0xffffffff813396e4 570 0 1531 879 75 [k] mem_cgroup_charge
25.81% 50.00% 0.00% 0.00% 0.00% 0x54 0 1 0xffffffff81331472 949 70 593 3359 74 [k] get_mem_cgroup_from_mm
19.35% 0.00% 0.00% 0.00% 0.00% 0x54 0 1 0xffffffff81339686 1352 0 1073 1022 74 [k] mem_cgroup_charge
9.68% 0.00% 0.00% 0.00% 0.00% 0x54 0 1 0xffffffff813396d6 1401 0 863 768 74 [k] mem_cgroup_charge
3.23% 0.00% 0.00% 0.00% 0.00% 0x54 0 1 0xffffffff81333106 618 0 804 11 9 [k] uncharge_batch
The offset 0x10 and 0x54 used to displayed in 2 groups, and now they are
listed together to give users a hint of extended false sharing.
[1]. https://lore.kernel.org/lkml/20201102091543.GM31092@shao2-debian/
Committer notes:
Link: https://lore.kernel.org/r/Y+wvVNWqXb70l4uy@feng-clx
Removed -a, leaving just as --double-cl, as this probably is not used so
frequently and perhaps will be even auto-detected if we manage to record
the MSR where this is configured.
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Feng Tang <feng.tang@intel.com>
Tested-by: Leo Yan <leo.yan@linaro.org>
Acked-by: Joe Mario <jmario@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tim Chen <tim.c.chen@intel.com>
Cc: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Link: https://lore.kernel.org/r/20230214075823.246414-1-feng.tang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The Retire Latency field is added in the var3_w of the
PERF_SAMPLE_WEIGHT_STRUCT. The Retire Latency reports pipeline stall of
this instruction compared to the previous instruction in cycles. That's
quite useful to display the information with perf mem report.
The p_stage_cyc for Power is also from the var3_w. Union the p_stage_cyc
and retire_lat to share the code.
Implement X86 specific codes to display the X86 specific header.
Add a new sort key retire_lat for the Retire Latency.
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20230104201349.1451191-8-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Sometimes users want to see actual (virtual) address of sampled instructions.
Add a new 'addr' sort key to display the raw addresses.
$ perf record -o- true | perf report -i- -s addr
# To display the perf.data header info, please use --header/--header-only options.
#
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.000 MB - ]
#
# Total Lost Samples: 0
#
# Samples: 12 of event 'cycles:u'
# Event count (approx.): 252512
#
# Overhead Address
# ........ ..................
#
42.96% 0x7f96f08443d7
29.55% 0x7f96f0859b50
14.76% 0x7f96f0852e02
8.30% 0x7f96f0855028
4.43% 0xffffffff8de01087
Note that it just compares and displays the sample ip. Each process can
have a different memory layout and the ip will be different even if they run
the same binary. So this sort key is mostly meaningful for per-process
profile data.
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20220923173142.805896-4-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Sort key 'p_stage_cyc' is used to present the latency cycles spent in
pipeline stages.
perf has local 'p_stage_cyc' sort key to display this info. There is no
global variant available for this sort key. The local variant shows
latency in a single sample, whereas the global value will be useful to
present the total latency (sum of latencies) in the hist entry. It
represents the latency number multiplied by the number of samples.
Add global ('p_stage_cyc') and local variant ('local_p_stage_cyc') for
this sort key. Use 'local_p_stage_cyc' as default option for "mem" sort
mode.
Also add this to the list of dynamic sort keys and made the
"dynamic_headers" and "arch_specific_sort_keys" as static.
Reported-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Athira Jajeev <atrajeev@linux.vnet.ibm.com>
Tested-by: Nageswara R Sastry <rnsastry@linux.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: linuxppc-dev@lists.ozlabs.org
Link: https://lore.kernel.org/r/20211203022038.48240-1-atrajeev@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
