New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 848499 link

Starred by 2 users

Issue metadata

Status: Available
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Fuchsia
Pri: 3
Type: Bug



Sign in to add a comment

WaitableEventPerfTest.Throughput consistently failing

Project Member Reported by stephanstross@google.com, May 31 2018

Issue description

The WaitableEventPerfTest.Throughput test in base:base_perftests fails consistently on Fuchsia and Linux

Linux Output:
[ RUN      ] WaitableEventPerfTest.Throughput
*RESULT counts: throughput= 333592 signals
*RESULT signal_time: throughput= 1501 ns/sample
*RESULT wait_time: throughput= 1971 ns/sample
../../base/synchronization/waitable_event_perftest.cc:174: Failure
Expected: (event.signal_times().capacity()) <= (kCapacity), actual: 4000000 vs 500000
Stack trace:
#0 0x0000003a32fc testing::internal::UnitTestImpl::CurrentOsStackTraceExceptTop()
#1 0x0000003a2cd9 testing::internal::AssertHelper::operator=()
#2 0x0000002da8ca base::WaitableEventPerfTest_Throughput_Test::TestBody()

[  FAILED  ] WaitableEventPerfTest.Throughput (1009 ms)

Fuchsia Output:
[ RUN      ] WaitableEventPerfTest.Throughput
*RESULT counts: throughput= 2475511 signals
*RESULT signal_time: throughput= 1054 ns/sample
*RESULT wait_time: throughput= 1358 ns/sample
../../base/synchronization/waitable_event_perftest.cc:174: Failure
Expected: (event.signal_times().capacity()) <= (kCapacity), actual: 4000000 vs 500000
Stack trace:
#00: testing::internal::UnitTestImpl::CurrentOsStackTraceExceptTop(int) at gtest.cc:?
#01: testing::internal::AssertHelper::operator=(testing::Message const&) const at gtest.cc:?
#02: base::WaitableEventPerfTest_Throughput_Test::TestBody() at waitable_event_perftest.cc:?

../../base/synchronization/waitable_event_perftest.cc:175: Failure
Expected: (event.wait_times().capacity()) <= (kCapacity), actual: 4000000 vs 500000
Stack trace:
#00: testing::internal::UnitTestImpl::CurrentOsStackTraceExceptTop(int) at gtest.cc:?
#01: testing::internal::AssertHelper::operator=(testing::Message const&) const at gtest.cc:?
#02: base::WaitableEventPerfTest_Throughput_Test::TestBody() at waitable_event_perftest.cc:?

[  FAILED  ] WaitableEventPerfTest.Throughput (1012 ms)

This failure can be replicated by executing the base_perftests target on either linux or fuchsia with
--gtest_filter=WaitableEventPerfTest.Throughput
 
Description: Show this description
Project Member

Comment 2 by bugdroid1@chromium.org, Nov 7

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/2af43e66d1952f61d0b4fc82b865f12724b6407e

commit 2af43e66d1952f61d0b4fc82b865f12724b6407e
Author: Sami Kyostila <skyostil@chromium.org>
Date: Wed Nov 07 01:40:49 2018

Speed up message loop perf test and disable failing waitable event perf test

In order to reduce cycle time on the performance waterfall, drop the
perf test measurement period from 30s to 5s.

Bug: 897751, 848499
Change-Id: I3cd921b327044f02767bc8a43a6e99e5a8a6f48f
Reviewed-on: https://chromium-review.googlesource.com/c/1315967
Reviewed-by: François Doray <fdoray@chromium.org>
Commit-Queue: Sami Kyöstilä <skyostil@chromium.org>
Cr-Commit-Position: refs/heads/master@{#605922}
[modify] https://crrev.com/2af43e66d1952f61d0b4fc82b865f12724b6407e/base/message_loop/message_loop_task_runner_perftest.cc
[modify] https://crrev.com/2af43e66d1952f61d0b4fc82b865f12724b6407e/base/synchronization/waitable_event_perftest.cc

Sign in to add a comment