diff options
| author | David Horstmann <david.horstmann@linaro.org> | 2019-09-02 14:29:54 +0100 |
|---|---|---|
| committer | David Horstmann <david.horstmann@linaro.org> | 2019-09-02 15:49:28 +0100 |
| commit | 356d1a5d65b0c21b6dce57aca842e73458f6873a (patch) | |
| tree | 9ac8b7a83e849bd0102f3288407fbf971ca9623f | |
| parent | c3f4acca9caf04ef8788355c2be11a3a68ace5ad (diff) | |
| download | art-testing-356d1a5d65b0c21b6dce57aca842e73458f6873a.tar.gz | |
Fix run.py result calculation when using the --dont-auto-calibrate option.
tools/benchmarks/run.py currently only records results from output lines
that end with 'per iteration'. These do not appear when the benchmarks
are run with the --dont-auto-calibrate option as no bench runner is used.
Instead, allow any line that starts with 'benchmarks/' to be parsed.
This prevents a crash that happens when run.py tries to calculate
the geomean while running with --dont-auto-calibrate.
Test: tools/benchmarks/run.py --iterations 1 --dont-auto-calibrate
Change-Id: I179a7c1b4a98288757bebcc721f1d4f6bd7b05f1
| -rwxr-xr-x | tools/benchmarks/run.py | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tools/benchmarks/run.py b/tools/benchmarks/run.py index 2a90f3a..c3ea63f 100755 --- a/tools/benchmarks/run.py +++ b/tools/benchmarks/run.py @@ -164,7 +164,7 @@ def RunBench(apk, classname, try: for line in outerr.rstrip().splitlines(): - if not line.endswith('per iteration'): + if not line.startswith('benchmarks/'): continue name = line.split(":")[0].rstrip() score = float(line.split(":")[1].strip().split(" ")[0].strip()) |