diff options
| author | Denis Nikitin <denik@google.com> | 2019-11-04 07:18:15 -0800 |
|---|---|---|
| committer | Denis Nikitin <denik@chromium.org> | 2019-11-05 06:27:52 +0000 |
| commit | fe8a44aa39ab93632e976b4aa5d00013601cdbc2 (patch) | |
| tree | d9323e74ee377436d1680df69f4f080be4f35ed0 /crosperf/schedv2.py | |
| parent | b1fc678ca17b224bfd7af33f96ab6b55eec47be0 (diff) | |
| download | toolchain-utils-fe8a44aa39ab93632e976b4aa5d00013601cdbc2.tar.gz | |
crosperf: Improve cooldown functionality
"Restart UI" procedure is changed.
DUT configuration including cooldown in SuiteRunner is now performed
while UI is off. Start UI after DUT configuration is done.
The change showed decrease of cooldown waiting time by a factor of 10.
Added Total experiment duration time in the report along with total
cooldown wait time.
Cooldown wait time is now printed per machine (with parallel execution
on multiple machines single cooldown time could be misleading).
To collect data per machine cooldown_time attribute is moved from
SuiteRunner to CrosMachine class.
BUG=chromium:966514
TEST=unittests and HW tests on kevin64 passed.
Change-Id: Ib3e145cd986ca7242d97b0aeeb7795ba066753fc
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/third_party/toolchain-utils/+/1897236
Tested-by: Denis Nikitin <denik@chromium.org>
Reviewed-by: George Burgess <gbiv@chromium.org>
Diffstat (limited to 'crosperf/schedv2.py')
| -rw-r--r-- | crosperf/schedv2.py | 10 |
1 files changed, 0 insertions, 10 deletions
diff --git a/crosperf/schedv2.py b/crosperf/schedv2.py index 768d29d8..dea3cdd3 100644 --- a/crosperf/schedv2.py +++ b/crosperf/schedv2.py @@ -205,8 +205,6 @@ class DutWorker(Thread): intel_pstate = self._sched.get_experiment().intel_pstate # Firstly, handle benchmarkruns that have cache hit. br = self._sched.get_cached_benchmark_run() - # Total wait time for cooling down. - total_waittime = 0 while br: try: self._stat_annotation = 'finishing cached {}'.format(br) @@ -243,17 +241,9 @@ class DutWorker(Thread): if self._kerncmd_update_needed(intel_pstate): self._update_kerncmd_intel_pstate(intel_pstate) - # When calculating cooldown wait time we assume that suite_runner is - # never reused so we can sum up the values across all benchmark_runs. - # If implementation changes causing the assert below to fail the - # calculation should be adjusted accordingly. - assert br.suite_runner.GetCooldownWaitTime() == 0 # Execute the br. self._execute_benchmark_run(br) - total_waittime += br.suite_runner.GetCooldownWaitTime() finally: - self._logger.LogOutput( - 'Total wait time for cooldown: %d min' % (total_waittime // 60)) self._stat_annotation = 'finished' # Thread finishes. Notify scheduler that I'm done. self._sched.dut_worker_finished(self) |