GitLab

IndexedRuleset.median_match_time_avg/MatchAll perftest shows a
regression after r759462 due to computation of rule count in the
UrlPatternIndexMatcher constructor. This CL changes
UrlPatternIndexMatcher rule count to be lazily computed which should
ensure that the cost of computation is only paid by clients which use
the rule count.

When the perf test is run locally on my machine:

Without this CL: *RESULT IndexedRuleset.median_match_time: MatchAll=
231364 us
With this CL: *RESULT IndexedRuleset.median_match_time:
MatchAll= 80248 us

Note that the test regression is exaggerated since the test creates a
UrlPatternIndexMatcher for each request.

BUG=1071599

Change-Id: I6364e4b9043494d18389a41c517b90756ba4c8f9
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2153693
Commit-Queue: Karan Bhatia <karandeepb@chromium.org>
Auto-Submit: Karan Bhatia <karandeepb@chromium.org>
Reviewed-by: Charlie Harrison <csharrison@chromium.org>
Cr-Commit-Position: refs/heads/master@{#759905}