android_kernel_xiaomi_sm8350/tools/perf
Frederic Weisbecker d6b17bebd7 perf: Provide a new deterministic events reordering algorithm
The current events reordering algorithm is based on a heuristic that
gets broken once we deal with a very fast flow of events.

Indeed the time period based flushing is not suitable anymore
in the following case, assuming we have a flush period of two
seconds.

    CPU 0           |        CPU 1
                    |
  cnt1 timestamps   |      cnt1 timestamps
                    |
    0               |         0
    1               |         1
    2               |         2
    3               |         3
    [...]           |        [...]
    4 seconds later

If we spend too much time to read the buffers (case of a lot of
events to record in each buffers or when we have a lot of CPU buffers
to read), in the next pass the CPU 0 buffer could contain a slice
of several seconds of events. We'll read them all and notice we've
reached the period to flush. In the above example we flush the first
half of the CPU 0 buffer, then we read the CPU 1 buffer where we
have events that were on the flush slice and then the reordering
fails.

It's simple to reproduce with:

	perf lock record perf bench sched messaging

To solve this, we use a new solution that doesn't rely on an
heuristical time slice period anymore but on a deterministic basis
based on how perf record does its job.

perf record saves the buffers through passes. A pass is a tour
on every buffers from every CPUs. This is made in order: for
each CPU we read the buffers of every counters. So the more
buffers we visit, the later will be the timstamps of their events.

When perf record finishes a pass it records a
PERF_RECORD_FINISHED_ROUND pseudo event.
We record the max timestamp t found in the pass n. Assuming these
timestamps are monotonic across cpus, we know that if a buffer
still has events with timestamps below t, they will be all available
and then read in the pass n + 1.
Hence when we start to read the pass n + 2, we can safely flush every
events with timestamps below t.

      ============ PASS n =================
         CPU 0         |   CPU 1
                       |
      cnt1 timestamps  |   cnt2 timestamps
            1          |         2
            2          |         3
            -          |         4  <--- max recorded

      ============ PASS n + 1 ==============
         CPU 0         |   CPU 1
                       |
      cnt1 timestamps  |   cnt2 timestamps
            3          |         5
            4          |         6
            5          |         7 <---- max recorded

        Flush every events below timestamp 4

      ============ PASS n + 2 ==============
         CPU 0         |   CPU 1
                       |
      cnt1 timestamps  |   cnt2 timestamps
            6          |         8
            7          |         9
            -          |         10

        Flush every events below timestamp 7
        etc...

It also works on perf.data versions that don't have
PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that
the events will be only flushed in the end of the perf.data
processing. It will then consume more memory and scale less with
large perf.data files.

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: Tom Zanussi <tzanussi@gmail.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-09 13:43:42 +02:00
..
arch perf probe: Add PowerPC DWARF register number mappings 2010-04-22 13:48:31 +10:00
bench perf: Fix endianness argument compatibility with OPT_BOOLEAN() and introduce OPT_INCR() 2010-04-14 11:26:44 +02:00
Documentation perf list: Improve the raw hw event descriptor documentation 2010-05-07 14:07:05 -03:00
scripts perf: Remove leftover useless options to record trace events from scripts 2010-04-30 19:55:00 +02:00
util perf: Provide a new deterministic events reordering algorithm 2010-05-09 13:43:42 +02:00
.gitignore perf: Ignore perf-archive temp file 2010-01-29 10:37:33 +01:00
builtin-annotate.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
builtin-bench.c perf bench: Add "all" pseudo subsystem and "all" pseudo suite 2009-12-14 08:51:19 +01:00
builtin-buildid-cache.c perf: Fix endianness argument compatibility with OPT_BOOLEAN() and introduce OPT_INCR() 2010-04-14 11:26:44 +02:00
builtin-buildid-list.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
builtin-diff.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
builtin-help.c perf: Fix endianness argument compatibility with OPT_BOOLEAN() and introduce OPT_INCR() 2010-04-14 11:26:44 +02:00
builtin-inject.c perf inject: Add missing bits 2010-05-04 10:48:22 -03:00
builtin-kmem.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
builtin-kvm.c perf: 'perf kvm' tool for monitoring guest performance from host 2010-04-19 12:37:24 +03:00
builtin-list.c perf list: Fix large list output by using the pager 2009-08-13 09:05:48 +02:00
builtin-lock.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
builtin-probe.c perf probe: Add --max-probes option 2010-04-26 15:35:20 -03:00
builtin-record.c perf: Introduce a new "round of buffers read" pseudo event 2010-05-09 13:43:42 +02:00
builtin-report.c perf report: Document '--call-graph' better for usage 2010-05-08 18:11:44 +02:00
builtin-sched.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
builtin-stat.c perf: Fix endianness argument compatibility with OPT_BOOLEAN() and introduce OPT_INCR() 2010-04-14 11:26:44 +02:00
builtin-test.c perf test: Initial regression testing command 2010-04-29 18:59:23 -03:00
builtin-timechart.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
builtin-top.c perf, x86: Improve the PEBS ABI 2010-05-07 11:31:02 +02:00
builtin-trace.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
builtin.h perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
command-list.txt perf inject: Add missing bits 2010-05-04 10:48:22 -03:00
CREDITS perf_counter tools: Add CREDITS file for Git contributors 2009-06-24 19:54:29 +02:00
design.txt perf: Fix few typos + cosmetics 2010-01-13 17:39:44 +01:00
Makefile perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
perf-archive.sh perf archive: Explain how to use the generated tarball 2010-03-23 20:33:08 +01:00
perf.c perf: add perf-inject builtin 2010-05-02 13:36:56 -03:00
perf.h perf: 'perf kvm' tool for monitoring guest performance from host 2010-04-19 12:37:24 +03:00