Bug 14602

Summary: JSC/WTF: fastMalloc doesn't use bsr under VC
Product: WebKit Reporter: 808caaa4.8ce9.9cd6c799e9f6
Component: JavaScriptCoreAssignee: Nobody <webkit-unassigned>
Status: RESOLVED WONTFIX    
Severity: Trivial CC: ap, barraclough
Priority: P4 Keywords: InRadar, PlatformOnly
Version: 523.x (Safari 3)   
Hardware: PC   
OS: Windows XP   

808caaa4.8ce9.9cd6c799e9f6
Reported 2007-07-13 04:08:35 PDT
With VC, fastMalloc.cpp doesn't use inline assembly for LgFloor(). #if defined(_MSC_VER) && !defined(_68K_) && !defined(_MPPC_) && !defined(_X86_) && !defined(_IA64_) && !defined(_AMD64_) && defined(_M_IX86) static inline int LgFloor(size_t n) { __asm xor eax,eax __asm not eax __asm bsr eax,n } #endif p.s. In WebKit.dll, do_malloc() and do_free() expands as ALWAYS_INLINE everywhere and costs binary sizes. Is this intended?
Attachments
David Kilzer (:ddkilzer)
Comment 1 2007-07-14 15:22:52 PDT
Maciej Stachowiak
Comment 2 2007-07-16 19:40:41 PDT
Yes, we don't have all the inline assembly implemented for Windows yet. It could be a small performance boost once done.
Maciej Stachowiak
Comment 3 2007-08-30 02:52:35 PDT
Why are the xor and the not needed in the inline assembly here? They don't seem to be needed in the gcc version.
Robert Blaut
Comment 4 2008-02-20 03:32:19 PST
P5 is not used for WebKit bugs. [http://webkit.org/quality/bugpriorities.html]
Gavin Barraclough
Comment 5 2012-09-06 12:50:49 PDT
I don't think this method has shown up under recent profiling, and this bug has sat idle for 5 years. I don't think this is a change we intend to make. Please reopen if you intend to work on this bug.
Note You need to log in before you can comment on or make changes to this bug.