Bug 142950 - ~13% of time spent in MarkRoots is unaccounted for in GC phase timings
Summary: ~13% of time spent in MarkRoots is unaccounted for in GC phase timings
Status: NEW
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: 528+ (Nightly build)
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Mark Hahnenberg
URL:
Keywords:
Depends on:
Blocks: 140774
  Show dependency tree / graph
 
Reported: 2015-03-22 13:08 PDT by Mark Hahnenberg
Modified: 2022-10-09 13:42 PDT (History)
6 users (show)

See Also:

Attachments
MarkRoots lots of "Other" (97.52 KB, image/png)
2015-03-22 19:25 PDT, Mark Hahnenberg 		no flags Details
Patch (6.31 KB, patch)
2015-03-28 10:16 PDT, Mark Hahnenberg 		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 Mark Hahnenberg 2015-03-22 13:08:45 PDT 
While messing around with graphing the time spent in each GC phase, I noticed that more than 13% of the time spent in the marking phase is currently unaccounted for (i.e. there is no active GCTimer for 13% of time spent in marking). There could be some GC pause time optimizations hiding in there, so we should figure out where that time is going.
Comment 1 Mark Hahnenberg 2015-03-22 19:25:44 PDT 
Created attachment 249216 [details]
MarkRoots lots of "Other"

Here's some graph output for the tool I mentioned. This run shows ~12% of "Other". Even though it says there are 9 categories in "Other", if I modify the tool so that those categories aren't included in "Other", it will still be ~12%.
Comment 2 Mark Hahnenberg 2015-03-28 10:16:06 PDT 
Created attachment 249663 [details]
Patch
Comment 3 Geoffrey Garen 2015-03-28 18:10:51 PDT 
Comment on attachment 249663 [details]
Patch

r=me
Comment 4 Ahmad Saleem 2022-10-09 06:23:30 PDT 
I checked via BugID and it seems that this r+ patch didn't landed yet.

Do we need it now? Thanks!