Bug 154086

Summary: [JSC] The destination of Sqrt should be Def, not UseDef
Product: WebKit Reporter: Benjamin Poulain <benjamin>
Component: New BugsAssignee: Benjamin Poulain <benjamin>
Status: RESOLVED FIXED    
Severity: Normal CC: commit-queue, fpizlo, keith_miller, mark.lam, msaboff, saam
Priority: P2    
Version: WebKit Nightly Build   
Hardware: Unspecified   
OS: Unspecified   
Attachments:
Description Flags
Patch none

Benjamin Poulain
Reported 2016-02-10 14:00:16 PST
[JSC] The destination of Sqrt should be Def, not UseDef
Attachments
Patch (1.43 KB, patch)
2016-02-10 14:01 PST, Benjamin Poulain 		no flags
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Benjamin Poulain
Comment 1 2016-02-10 14:01:09 PST
Created attachment 271025 [details] Patch
Geoffrey Garen
Comment 2 2016-02-10 14:22:21 PST
Comment on attachment 271025 [details] Patch r=me
WebKit Commit Bot
Comment 3 2016-02-10 17:35:43 PST
Comment on attachment 271025 [details] Patch Clearing flags on attachment: 271025 Committed r196409: <http://trac.webkit.org/changeset/196409>
WebKit Commit Bot
Comment 4 2016-02-10 17:35:46 PST
All reviewed patches have been landed. Closing bug.
Filip Pizlo
Comment 5 2016-02-11 13:26:36 PST
Did you test performance? The last time we did a fix like this it affected performance adversely because we forgot some other stuff. It looks like SqrtFloat doesn't have an entry in FixPartialRegisterStalls so this could be an issue here.
Benjamin Poulain
Comment 6 2016-02-11 14:01:48 PST
(In reply to comment #5) > Did you test performance? > > The last time we did a fix like this it affected performance adversely > because we forgot some other stuff. It looks like SqrtFloat doesn't have an > entry in FixPartialRegisterStalls so this could be an issue here. This is a subpart of the aliasing patch, the perf was fine when I did it (2 kraken tests improve a tiny bit but that was it).
Note You need to log in before you can comment on or make changes to this bug.