[Pixman] Latest GIT source for 'pixman-sse2.c'

[Matt Turner]

On Sun, Oct 6, 2013 at 1:50 AM, John Emmas <johne53 at tiscali.co.uk> wrote:
> On 05/10/2013 19:32, John Emmas wrote:
>>
>> On 5 Oct 2013, at 19:00, Siarhei Siamashka wrote:
>>
>>> Andrea Canciani has already investigated the problem and submitted the
>>> fixes here:
>>>
>>>
>>> http://lists.freedesktop.org/archives/pixman/2013-September/002954.html
>>>
>> Many thanks for the super fast response guys.  I'm at a different PC now
>> but I'll apply that patch tomorrow.
>
>
> I applied that patch this morning and sure enough, it does fix the problem.
> Thanks to Andrea for noticing it.
>
> BTW...  while reading the patch I noticed that, quite by accident, the
> source file 'pixman-mmx.c' had somehow gotten excluded from my MSVC build
> project, so I took the opportunity to add it.  Although the build still
> succeeds, I see several warnings of this form while building
> 'pixman-mmx.c':-
>
>       pixman-mmx.c(586) : warning C4799: function 'whatever' has no EMMS
> instruction
>
> I don't know if that means anything bad but I thought it wouldn't do any
> harm flag it up.  Here's a list of the affected functions:-
>
>       function 'expand_4xpacked565' has no EMMS instruction
>       function 'is_opaque' has no EMMS instruction
>       function 'is_equal' has no EMMS instruction
>       function 'to_uint64' has no EMMS instruction
>       function 'expand_4x565' has no EMMS instruction
>       function 'is_zero' has no EMMS instruction
>       function 'store8888' has no EMMS instruction

All of these are all inline functions, so _mm_empty() isn't required.

>       function 'fast_composite_scaled_bilinear_mmx_8888_8_8888_none_OVER'
> has no EMMS instruction
>       function 'fast_composite_scaled_bilinear_mmx_8888_8_8888_pad_OVER' has
> no EMMS instruction

This has _mm_empty().