601016 - JM: Add detection for all levels of SSE support

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Alex Miller]

User-Agent:       Mozilla/5.0 (X11; U; Linux x86_64; en-US) AppleWebKit/533.4 (KHTML, like Gecko) Chrome/5.0.375.127 Safari/533.4
Build Identifier: 

As there's currently only SSE2, SSE4.1 and SSE4.2 detection available, it made sense to finish it off and add support for checking all possible levels.

Reproducible: Always

Comment 1

[Alex Miller]

Attachment: patch v1

Adds checking for all levels of SSE. The check to see if CPUID exists isn't done... I'm not even aware of a processor that doesn't support it anymore.

Comment 2

[Sean Stangl [:sstangl]]

Evidently SSE detection never worked on x64 -- it was fine because we never called the check function.

Changing jscntxt.cpp:2201 (updateJITEnabled()) to
> # if defined JS_CPU_X86 || defined JS_CPU_X64

to force getSSEState() to be called results in a build error. We should be able to check SSE state even on x64, even though AMD64 implies >= SSE2.

Comment 3

[Alex Miller]

Attachment: x64 fix v1

Copied the existing version for gcc, changed the pushl/popl to pushq/popq, and threw a #if WTF_CPU_X86_64 around it.

Comment 4

[Sean Stangl [:sstangl]]

http://hg.mozilla.org/tracemonkey/rev/f8a4a5f68942

Comment 5

[Makoto Kato [:m_kato]]

This check in causes bugstage on Win64

LINK : warning LNK4098: defaultlib 'MSVCRTD' conflicts with use of other libs; use /NODEFAULTLIB:library
jscntxt.obj : error LNK2019: unresolved external symbol "void __cdecl __cpuid(int * const,int)" (?__cpuid@@YAXQEAHH@Z) referenced in function "private: static void __cdecl JSC::MacroAssemblerX86Common::setSSECheckState(void)" (?setSSECheckS
tate@MacroAssemblerX86Common@JSC@@CAXXZ)
mozjs.dll : fatal error LNK1120: 1 unresolved externals

Comment 6

[Makoto Kato [:m_kato]]

Attachment: fix for win64

Comment 7

[Sean Stangl [:sstangl]]

Comment on attachment 480024 [details] [diff] [review]
fix for win64

Thanks for catching this. This fix should be upstreamed at some point, since the problem exists in MASM as exists in the WebKit tree.

Comment 8

[Sean Stangl [:sstangl]]

http://hg.mozilla.org/tracemonkey/rev/e7359994523f

Comment 9

[Robert Sayre]

http://hg.mozilla.org/mozilla-central/rev/e7359994523f