Bug 70588 - Bencher script doesn't measure GC times accurately
Summary: Bencher script doesn't measure GC times accurately
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Tools / Tests (show other bugs)
Version: 528+ (Nightly build)
Hardware: All All
: P2 Normal
Assignee: Nobody
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-10-21 01:32 PDT by Filip Pizlo
Modified: 2011-10-21 16:20 PDT (History)
0 users

See Also:

Attachments
the patch (10.92 KB, patch)
2011-10-21 01:41 PDT, Filip Pizlo 		ggaren: review+
Details | Formatted Diff | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Filip Pizlo 2011-10-21 01:32:18 PDT 
The bencher script runs benchmarks using the SunSpider harness style, where gc() is invoked between samples.  This means that GC times tend to not be measured.  But particularly SunSpider and V8 have significant GC pressure when run repeatedly.  The bencher script should retain its default mode (since it's great for pinpointing performance changes in the mutator) but have options to also measure GC times.
Comment 1 Filip Pizlo 2011-10-21 01:41:45 PDT 
Created attachment 111921 [details]
the patch
Comment 2 Geoffrey Garen 2011-10-21 12:06:52 PDT 
Comment on attachment 111921 [details]
the patch

rs=me
Comment 3 Filip Pizlo 2011-10-21 16:20:01 PDT 
Landed in http://trac.webkit.org/changeset/98167