233a0bb6d6
Set precise_ip to 0 for cache on Intel.
...
Fuck knows how this is supposed to work. perf_event_open() opens the
descriptor successfully, but it produces no samples, if precise_ip is not 0.
There are no such problems on ARM (where precise_ip is 3, but maybe it is not
supported at all on that architecture, again, fuck knows if), and on AMD
perf_event_open() does not succeed when precise_ip > 0.
2021-05-23 19:45:13 +02:00
b2d5fe8e1f
Reduce sampling frequency.
2021-05-23 19:03:11 +02:00
b1e4d16537
PIDs are no longer needed in samples.
2021-05-23 19:00:45 +02:00
bbd1c4505c
Restrict perf to return events only for the current PID.
2021-05-23 18:53:09 +02:00
4ad6f682c8
Exclude VM-related stuff.
2021-05-23 18:44:16 +02:00
fece23a32b
Set frequency, not period.
...
This enables sampling on ARM dev board.
2021-05-23 18:02:06 +02:00
7d3119cbac
Remove irrelevant flag.
2021-05-23 18:01:18 +02:00
34ca6d865e
Sample branches and cache more frequently.
2021-05-22 02:28:32 +02:00
c7026cbc1f
Reduce hw sample period.
2021-05-22 02:27:34 +02:00
089eda0be9
Precise_ip should be shared in each pair of counters.
2021-05-22 02:16:49 +02:00
fef507dfa2
Merge remote-tracking branch 'origin/master' into hw
2021-05-22 02:05:47 +02:00
68948712b4
Don't sleep if queues are empty, but there's queries to handle.
2021-05-22 01:12:42 +02:00
1e6aedf9e6
Limit client query response rate.
...
Original idea by xavier <xavierb@gmail.com >
2021-05-22 01:05:06 +02:00
d7541bbdba
Allow disabling inline resolution on windows.
...
Original commit a6b25497 by xavier <xavierb@gmail.com >:
add TRACY_CALLSTACK_IGNORE_INLINES to tradeoff speed vs precision in win32 DecodeCallstackPtr()
SymQueryInlineTrace() is too slow in some cases:
300000 queries backlog getting processed at ~70 per second is prohibitive.
(without inlines resolution, it's more like ~20000 queries per second)
2021-05-21 22:27:35 +02:00
8ec08465ee
Add debug messages to perf event setup.
2021-05-21 01:47:45 +02:00
afcebb6e6a
Add debug print macros.
2021-05-21 01:47:31 +02:00
4d668741eb
Probe for acceptable precise_ip value.
...
This is stupid, but it's exactly what perf does... Sigh.
2021-05-21 01:33:37 +02:00
bcb7b94272
Tid is not needed.
2021-05-20 02:39:22 +02:00
f0f3babacf
Set correct message types.
2021-05-20 02:27:36 +02:00
5f3d1c0faf
Sample cache and branch stats.
2021-05-20 02:15:23 +02:00
faf87809d7
Reduce hw sampling rate.
2021-05-20 01:48:52 +02:00
741de5c8fb
Allow disabling cycle/retirement sampling.
2021-05-19 23:38:32 +02:00
2e38e70049
Reduce hardware sampling perdiod. Don't sample time.
2021-05-19 23:21:21 +02:00
101cdd9b4b
Don't send thread id for hw samples.
2021-05-19 22:52:13 +02:00
7794443453
Collect CPU cycles and instruction retirement events.
2021-05-19 21:09:55 +02:00
16101571e0
Close perf_event file descriptor on exec.
2021-05-19 21:09:55 +02:00
9cd1b26bc7
Keep count of ring buffers separate from number of CPUs.
2021-05-19 21:09:55 +02:00
b7d52d2eab
Store RingBuffer identifier.
2021-05-19 21:09:52 +02:00
42a272edf5
Allow control of sampling frequency.
2021-05-11 18:31:20 +02:00
a6c6943a6c
Check if GetThreadDescription() is supported.
...
This functionality is available since Win 10 1607.
2021-05-04 16:13:42 +02:00
eb7d220eea
Added support for TRACY_NO_FRAME_IMAGE define.
2021-04-29 20:55:16 +02:00
56f0bdd571
ARM doesn't follow x64 canonical address requirements.
2021-04-29 18:24:37 +02:00
505656df5a
Trace frame count may be zero.
2021-04-29 18:24:37 +02:00
915693ac39
Use tracy_malloc rather than 'new' in ProfilerThreadDataKey
...
This codepath, involving a workaround for GCC < 8.4, called 'new' and
'delete' directly, which could cause infinite recursion when
user-provided versions of those functions were themselves using Tracy
functionality.
Now, this codepath uses Tracy's internal allocator.
See issues #194 , #196
2021-04-12 10:06:35 -07:00
40efbe8529
Use rpmalloc for initialization-related allocations.
2021-04-10 13:02:32 +02:00
2bb5d126fd
rpmalloc_thread_initialize is called in RPMallocInit.
2021-04-10 12:55:00 +02:00
fa942d18fe
Fix spacing
2021-04-09 15:35:44 -07:00
e845c23493
Removed duplicate function
2021-04-09 15:35:07 -07:00
3fad55d7bc
Missed one
2021-04-09 15:34:21 -07:00
eac23cead2
PR feedback
2021-04-09 15:33:01 -07:00
fc142b4f9c
fixed a build break on AARCH64.
2021-04-09 13:50:35 -07:00
00ac6d1d8e
worked around Windows broken getenv() call.
2021-04-09 13:50:02 -07:00
76a02205f3
Fix centos shutdown crash
2021-04-09 11:58:34 -07:00
c288a7903b
Make {Startup,Shutdown}Profiler() signatures consistent.
2021-03-08 02:39:51 +01:00
99c6b91c0c
Fix sending GPU context name in on-demand mode.
2021-02-27 19:59:32 +01:00
c12de1b326
Merge pull request #178 from sideeffects/master
...
Add IsActive accessor to ScopedZone.
2021-02-16 20:52:29 +01:00
29af8352ee
Add IsActive accessor to ScopedZone.
2021-02-12 20:30:43 -06:00
5ea71ea20d
Apparently program_invocation_short_name may be not defined.
2021-02-11 18:12:59 +01:00
26a8ec3909
Reuse existing variable.
2021-02-10 18:56:07 +01:00
5e48eebf26
Fix type in comparison.
2021-02-07 21:08:24 +01:00
Bartosz Taudul
Bartosz Taudul
JW
joshuakr
Eric van Beurden
Joshua Kriegshauser
John Lynch