Bug 144205 - [JSC] Implement Math.clz32(), remove Number.clz()
Summary: [JSC] Implement Math.clz32(), remove Number.clz()
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: 528+ (Nightly build)
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Benjamin Poulain
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-04-25 19:35 PDT by Benjamin Poulain
Modified: 2015-04-26 22:04 PDT (History)
2 users (show)

See Also:

Attachments
Patch (48.97 KB, patch)
2015-04-25 19:47 PDT, Benjamin Poulain 		no flags 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 Benjamin Poulain 2015-04-25 19:35:54 PDT 
[JSC] Implement Math.clz32(), remove Number.clz()
Comment 1 Benjamin Poulain 2015-04-25 19:47:49 PDT 
Created attachment 251664 [details]
Patch
Comment 2 Michael Saboff 2015-04-26 08:15:41 PDT 
Comment on attachment 251664 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=251664&action=review

r=me with one question.

> Source/JavaScriptCore/assembler/MacroAssemblerX86Common.h:194
> +        xor32(TrustedImm32(0x1f), dst);

Why is the xor needed?  Since you protect for the zero case and bsr is spec'ed to return an unsigned value from 0..31, you shouldn't need this xor.
Comment 3 Benjamin Poulain 2015-04-26 11:15:58 PDT 
(In reply to comment #2)
> > Source/JavaScriptCore/assembler/MacroAssemblerX86Common.h:194
> > +        xor32(TrustedImm32(0x1f), dst);
> 
> Why is the xor needed?  Since you protect for the zero case and bsr is
> spec'ed to return an unsigned value from 0..31, you shouldn't need this xor.

BSR returns the position of the most significant bit, the opposite of what we need.
The xor32 is there to inverse the result.
Comment 4 Benjamin Poulain 2015-04-26 12:56:28 PDT 
Comment on attachment 251664 [details]
Patch

Clearing flags on attachment: 251664

Committed r183358: <http://trac.webkit.org/changeset/183358>
Comment 5 Benjamin Poulain 2015-04-26 12:56:31 PDT 
All reviewed patches have been landed.  Closing bug.
Comment 6 Michael Saboff 2015-04-26 22:04:16 PDT 
(In reply to comment #3)
> (In reply to comment #2)
> > > Source/JavaScriptCore/assembler/MacroAssemblerX86Common.h:194
> > > +        xor32(TrustedImm32(0x1f), dst);
> > 
> > Why is the xor needed?  Since you protect for the zero case and bsr is
> > spec'ed to return an unsigned value from 0..31, you shouldn't need this xor.
> 
> BSR returns the position of the most significant bit, the opposite of what
> we need.
> The xor32 is there to inverse the result.

Got it.