Add trace rerun options to benchmark and pass them through chartjson.
After discussion with Annie/Dave/Fadi/Kari and myself, seems like we settled on not exposing an arbitrary # of new command lines. Rather, there'd be an optional param to rerun with extra tracing categories enabled. The benchmark would specify that there's a debug option available, and has a way of customizing the exact categories it needs.
Example of v8 benchmark modified:
class V8GarbageCollectionCases(benchmark.Benchmark):
"""Measure V8 metrics on the garbage collection cases."""
def CreateTimelineBasedMeasurementOptions(self):
# TODO(ernstm): Remove v8-overhead when benchmark relevant v8 events become
# available in the 'benchmark' category.
# This can theoretically be handled now by the "default" override
return timeline_based_measurement.Options()
@classmethod
def HasBenchmarkTraceRerunDebugOption(cls):
return True
def SetupBenchmarkDefaultTraceRerunOptions(self, tbm_options):
tbm_options.ExtendTraceCategoryFilters(['v8'])
def SetupBenchmarkDebugTraceRerunOptions(self, tbm_options):
tbm_options.ExtendTraceCategoryFilters(['*'])
BUG=
Review URL: https://codereview.chromium.org/840043004
Cr-Commit-Position: refs/heads/master@{#313080}
3 files changed
