[PATCHv5 00/22] perf stat: Introduce --per-thread option

[Jiri Olsa <jolsa@kernel.org>]

hi,
adding the possibility to display stat data per thread.

Allowing following commands and output:

  $ perf stat  -e cycles,instructions --per-thread -p 30190,30242
  ^C
   Performance counter stats for process id '30190,30242':

               cat-30190                     0      cycles
               yes-30242         3,842,525,421      cycles
               cat-30190                     0      instructions
               yes-30242        10,370,817,010      instructions

         1.143155657 seconds time elapsed


Also works under interval mode:

  $ perf stat  -e cycles,instructions --per-thread -p 30190,30242 -I 1000
  #           time             comm-pid                  counts unit events
       1.000073435              cat-30190                89,058      cycles
       1.000073435              yes-30242         3,360,786,902      cycles                     (100.00%)
       1.000073435              cat-30190                14,066      instructions
       1.000073435              yes-30242         9,069,937,462      instructions
       2.000204830              cat-30190                     0      cycles
       2.000204830              yes-30242         3,351,667,626      cycles
       2.000204830              cat-30190                     0      instructions
       2.000204830              yes-30242         9,045,796,885      instructions
  ^C     2.771286639              cat-30190                     0      cycles
       2.771286639              yes-30242         2,593,884,166      cycles
       2.771286639              cat-30190                     0      instructions
       2.771286639              yes-30242         7,001,171,191      instructions


Available in here:
  git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
  perf/per_thread

v5 changes:
  - reading comms explicitly by thread_map__read_comms,
    so it gets read only within 'perf stat --per-thread'
  - introducing thread_map__reset function
  - several patches got accepted, rebased on latest perf/core

v4 changes:
  - add thread_map__set_(pid|comm) interface
  - several fixes for task comm string reading
  - added perf_evlist__reset_stats function

v3 changes:
  - some patches already pulled in, rebased on latest acme's perf/core
  - fixed cpu_map__get/thread_map__get to accept NULL map
  - renamed propagate_maps to perf_evlist__propagate_maps
  - renamed --per-task option to --per-thread
  - changed 'task-pid' header output to 'comm-pid'

v2 changes:
  - renamed xyarray__zero to xyarray__reset
  - using atomic_t for cpu_map/thread_map refcnt
  - moving all stat functions into stat object
  - using init_stats to initialize 'struct stats*'
  - change TASK-$pid output into $comm-$pid
  - updated doc with --per-task option


thanks,
jirka
---
Jiri Olsa (22):
      perf thread_map: Introduce thread_map__reset function
      perf tools: Add comm string into struct thread_map
      perf tests: Add thread_map object tests
      perf stat: Introduce perf_counts function
      perf stat: Use xyarray for cpu evsel counts
      perf stat: Make stats work over the thread dimension
      perf stat: Rename struct perf_counts::cpu member to values
      perf stat: Introduce perf_evlist__reset_stats
      perf stat: Move perf_evsel__(alloc|free|reset)_stat_priv into stat object
      perf stat: Move perf_evsel__(alloc|free)_prev_raw_counts into stat object
      perf stat: Move perf_evlist__(alloc|free|reset)_stats into stat object
      perf stat: Introduce perf_evsel__alloc_stats function
      perf stat: Introduce perf_evsel__read function
      perf stat: Introduce read_counters function
      perf stat: Separate counters reading and processing
      perf stat: Move zero_per_pkg into counter process code
      perf stat: Move perf_stat initialization counter process code
      perf stat: Remove perf_evsel__read_cb function
      perf stat: Rename print_interval to process_interval
      perf stat: Using init_stats instead of memset
      perf stat: Introduce print_counters function
      perf stat: Introduce --per-thread option

 tools/perf/Documentation/perf-stat.txt     |   4 +
 tools/perf/builtin-stat.c                  | 402 +++++++++++++++++++++++++++++++------------------------------
 tools/perf/tests/Build                     |   1 +
 tools/perf/tests/builtin-test.c            |   4 +
 tools/perf/tests/openat-syscall-all-cpus.c |   6 +-
 tools/perf/tests/openat-syscall.c          |   4 +-
 tools/perf/tests/tests.h                   |   1 +
 tools/perf/tests/thread-map.c              |  38 ++++++
 tools/perf/util/evlist.h                   |   1 -
 tools/perf/util/evsel.c                    |  24 ++--
 tools/perf/util/evsel.h                    |  28 +----
 tools/perf/util/python-ext-sources         |   1 +
 tools/perf/util/stat.c                     | 132 ++++++++++++++++++--
 tools/perf/util/stat.h                     |  47 +++++++-
 tools/perf/util/thread_map.c               |  76 +++++++++++-
 tools/perf/util/thread_map.h               |   8 ++
 16 files changed, 523 insertions(+), 254 deletions(-)
 create mode 100644 tools/perf/tests/thread-map.c