Exclude noisy benchmarks from labelling computations (#29039)

* Exclude noisy benchmarks from labelling computations

* Automated change: Fix sanity tests

Co-authored-by: ctiller <ctiller@users.noreply.github.com>
pull/29019/head
Craig Tiller 3 years ago committed by GitHub
parent 00d587dc56
commit 40adbb46f0
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
  1. 6
      tools/profiling/microbenchmarks/bm_diff/bm_diff.py

@ -214,7 +214,11 @@ def diff(bms, loops, regex, track, old, new, counters):
# figure out the significance of the changes... right now we take the 95%-ile
# benchmark delta %-age, and then apply some hand chosen thresholds
histogram = []
for bm in benchmarks.values():
_NOISY = ["BM_WellFlushed"]
for name, bm in benchmarks.items():
if name in _NOISY:
print("skipping noisy benchmark '%s' for labelling evaluation" %
name)
if bm.skip():
continue
d = bm.speedup['cpu_time']

Loading…
Cancel
Save