DescriptionEnable more benchmark_smoke_unittest coverage
Previously, benchmark_smoke_unittest has a heuristic to pick
a benchmark for smoke testing if the benchmark uses a test
in measurement directory.
This patch removes that check, and instead provide a black
list of benchmarks module that we exclude the
benchmark_smoke_unittest to avoid high CQ time.
This helped increases benchmark smoke coverage from 14 benchmarks -> 33 benchmarks (one in each benchmark module).
On my local linux machine, this increases the cycle time of
benchmark_smoke_unittest from 20s -> 1m7s.
BUG=490130
CQ_EXTRA_TRYBOTS=tryserver.chromium.perf:linux_perf_bisect;tryserver.chromium.perf:mac_perf_bisect;tryserver.chromium.perf:android_nexus5_perf_bisect
Committed: https://crrev.com/65014162e48a7f5881e8b282dd9e8dc17c5aa7c4
Cr-Commit-Position: refs/heads/master@{#330909}
Patch Set 1 : #Patch Set 2 : Blacklist rasterize_and_record_micro #Patch Set 3 : Black list blink_perf #Patch Set 4 : Blacklist dom_perf instead of blink_perf #
Messages
Total messages: 21 (11 generated)
|