Bug 164869 - Add support for rotate in B3 and the relevant assemblers
Summary: Add support for rotate in B3 and the relevant assemblers
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Keith Miller
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-11-17 08:57 PST by Keith Miller
Modified: 2016-11-17 09:52 PST (History)
4 users (show)

See Also:

Attachments
Patch (34.48 KB, patch)
2016-11-17 09:07 PST, Keith Miller 		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 Keith Miller 2016-11-17 08:57:42 PST 
Add support for rotate in B3 and the relevant assemblers
Comment 1 Keith Miller 2016-11-17 09:07:25 PST 
Created attachment 295051 [details]
Patch
Comment 2 Geoffrey Garen 2016-11-17 09:34:39 PST 
Comment on attachment 295051 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=295051&action=review

r=me

> Source/JavaScriptCore/assembler/MacroAssemblerARM64.h:670
> +    void rotateRight32(RegisterID src, TrustedImm32 imm, RegisterID dest)
> +    {
> +        m_assembler.ror<32>(dest, src, imm.m_value & 31);
> +    }

It looks like this would transform an identity rotation into a non-identity rotation. Is that desirable?
Comment 3 Mark Lam 2016-11-17 09:35:46 PST 
Comment on attachment 295051 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=295051&action=review

> Source/JavaScriptCore/ChangeLog:14
> +        ammount by the value. Additionally, this patch refactors some

/ammount/amount/
Comment 4 Keith Miller 2016-11-17 09:44:15 PST 
Comment on attachment 295051 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=295051&action=review

>> Source/JavaScriptCore/assembler/MacroAssemblerARM64.h:670
>> +    }
> 
> It looks like this would transform an identity rotation into a non-identity rotation. Is that desirable?

AFAIK, ARM doesn't specify what happens when the immediate is > 31. This matches the behavior if the immediate was put into a register.
Comment 5 Keith Miller 2016-11-17 09:44:28 PST 
Comment on attachment 295051 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=295051&action=review

>> Source/JavaScriptCore/ChangeLog:14
>> +        ammount by the value. Additionally, this patch refactors some
> 
> /ammount/amount/

fixed.
Comment 6 Keith Miller 2016-11-17 09:52:11 PST 
Committed r208848: <http://trac.webkit.org/changeset/208848>