| Index: tools/perf/benchmarks/memory.py
|
| diff --git a/tools/perf/benchmarks/memory.py b/tools/perf/benchmarks/memory.py
|
| index 4eb83b900ad65d963929ca45e0ba5d00020d6a87..b0eafacd7d72391ae634b955d433800967d4f0de 100644
|
| --- a/tools/perf/benchmarks/memory.py
|
| +++ b/tools/perf/benchmarks/memory.py
|
| @@ -41,6 +41,15 @@ class _MemoryInfra(perf_benchmark.PerfBenchmark):
|
| chrome_trace_config.MemoryDumpConfig())
|
| return tbm_options
|
|
|
| + def SetExtraBrowserOptions(self, options):
|
| + # Just before we measure memory we flush the system caches
|
| + # unfortunately this doesn't immediately take effect, instead
|
| + # the next page run is effected. Due to this the first page run
|
| + # has anomalous results. This option causes us to flush caches
|
| + # each time before Chrome starts so we effect even the first page
|
| + # - avoiding the bug.
|
| + options.clear_sytem_cache_for_browser_and_profile_on_start = True
|
| +
|
|
|
| # TODO(bashi): Workaround for http://crbug.com/532075.
|
| # @benchmark.Enabled('android') shouldn't be needed.
|
|
|