[Webkit-unassigned] [Bug 55046] Share VFP detection for ARM in MacroAssemblerARMCommon class

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=55046





--- Comment #10 from Xan Lopez <xan.lopez at gmail.com>  2011-07-13 05:46:19 PST ---
(In reply to comment #9)
> (From update of attachment 100105 [details])
> Hi xan, this patch is going to add unnecessary dynamic checks on GGC/__VFP_FP__ and RVCT/__TARGET_FPU_VFP platforms, so I'd like to suggest the following:
> 
> 1) remove s_isVFPPresent for non-LINUX platforms.
> 
> 2) define isVFPPresent() for non-LINUX platforms as ALWAYS_INLINE & returning a constant true/false, so that these checks can be compile out.
> 
> 3) for linux only, define isVFPPresent() to return the value of s_isVFPPresent.
> 
> 
> Hope this sounds good.

Hrm, I think I'm confused. s_isVFPPresent is the return value of isVFPPresent() already (and it's only called once to initialize it). So making isVFPPresent() be the value of s_isVFPPresent introduces a recursive dependency? :)

Also I'm basically just moving code around, so I suppose that whatever issue exists would be already there? In any case I didn't really understand the proposal. s_isVFPPresent is what's used in the MacroAssembler methods, and we use isVFPPresent() once to set the value. On Linux we use the /proc thing, and in the rest of platforms we just return a constant decided at compile time. But in any case it's only done once, unless I'm mistaken, and from there on s_isVFPPresent is used.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.