Add SSE4 check to BlurImage optimization.

Add SSE4 check to BlurImage optimization.

Adds a build-time SSE4 check to SkBlurImage_opts_SSE2.cpp
in the SkBoxBlur_SSE2 function.

Signed-off-by: Henrik Smiding <henrik.smiding@intel.com>;

Committed: http://code.google.com/p/skia/source/detail?r=14750

[mtklein, 2014-05-14 17:15:56]

Duh.  Good find!

https://codereview.chromium.org/281963002/diff/1/src/opts/SkBlurImage_opts_SS...
File src/opts/SkBlurImage_opts_SSE2.cpp (right):

https://codereview.chromium.org/281963002/diff/1/src/opts/SkBlurImage_opts_SS...
src/opts/SkBlurImage_opts_SSE2.cpp:58: #if SK_CPU_SSE_LEVEL >=
SK_CPU_SSE_LEVEL_SSE42
I think the comment below is correct, right?  Only needs SK_CPU_SSE_LEVEL_SSE41?

(My reference is https://software.intel.com/sites/landingpage/IntrinsicsGuide)

[Stephen White, 2014-05-14 21:56:46]

On 2014/05/14 17:15:56, mtklein wrote:
> Duh.  Good find!
> 
>
https://codereview.chromium.org/281963002/diff/1/src/opts/SkBlurImage_opts_SS...
> File src/opts/SkBlurImage_opts_SSE2.cpp (right):
> 
>
https://codereview.chromium.org/281963002/diff/1/src/opts/SkBlurImage_opts_SS...
> src/opts/SkBlurImage_opts_SSE2.cpp:58: #if SK_CPU_SSE_LEVEL >=
> SK_CPU_SSE_LEVEL_SSE42
> I think the comment below is correct, right?  Only needs
SK_CPU_SSE_LEVEL_SSE41?
> 
> (My reference is https://software.intel.com/sites/landingpage/IntrinsicsGuide)

In addition to a compile-time guard, I think this also needs to be protected by
a runtime CPUID check.

in fact, we'll need to generate two code paths here (maybe templated?), and call
the appropriate one based on
SSE level.

[henrik.smiding, 2014-05-15 15:11:54]

On 2014/05/14 21:56:46, Stephen White wrote:
> On 2014/05/14 17:15:56, mtklein wrote:
> > Duh.  Good find!
> > 

I added partial SSE4-support just recently, so it wasn't possible to change
until now.

> >
>
https://codereview.chromium.org/281963002/diff/1/src/opts/SkBlurImage_opts_SS...
> > File src/opts/SkBlurImage_opts_SSE2.cpp (right):
> > 
> >
>
https://codereview.chromium.org/281963002/diff/1/src/opts/SkBlurImage_opts_SS...
> > src/opts/SkBlurImage_opts_SSE2.cpp:58: #if SK_CPU_SSE_LEVEL >=
> > SK_CPU_SSE_LEVEL_SSE42
> > I think the comment below is correct, right?  Only needs
> SK_CPU_SSE_LEVEL_SSE41?
> > 
> > (My reference is
https://software.intel.com/sites/landingpage/IntrinsicsGuide)
> 
> In addition to a compile-time guard, I think this also needs to be protected
by
> a runtime CPUID check.
> 
> in fact, we'll need to generate two code paths here (maybe templated?), and
call
> the appropriate one based on
> SSE level.

Yes, but we also need a new static lib build with SSE4x enabled, before I can do
the 'ultimate' version.
I have the framework for this in another issue/patch. Once I get that merged
I'll fix this.

However, the current version should still work in a limited scope for Android
framework builds.

[henrik.smiding, 2014-05-15 15:12:05]

https://codereview.chromium.org/281963002/diff/1/src/opts/SkBlurImage_opts_SS...
File src/opts/SkBlurImage_opts_SSE2.cpp (right):

https://codereview.chromium.org/281963002/diff/1/src/opts/SkBlurImage_opts_SS...
src/opts/SkBlurImage_opts_SSE2.cpp:58: #if SK_CPU_SSE_LEVEL >=
SK_CPU_SSE_LEVEL_SSE42
On 2014/05/14 17:15:56, mtklein wrote:
> I think the comment below is correct, right?  Only needs
SK_CPU_SSE_LEVEL_SSE41?
> 
> (My reference is https://software.intel.com/sites/landingpage/IntrinsicsGuide)

Done.

[mtklein, 2014-05-15 15:20:10]

> Yes, but we also need a new static lib build with SSE4x enabled, before I can
do
> the 'ultimate' version.
> I have the framework for this in another issue/patch. Once I get that merged
> I'll fix this.
> 
> However, the current version should still work in a limited scope for Android
> framework builds.

sgtm + lgtm.

[henrik.smiding, 2014-05-15 15:53:38]

The CQ bit was checked by henrik.smiding@intel.com

[commit-bot: I haz the power, 2014-05-15 15:54:00]

CQ is trying da patch. Follow status at

https://skia-tree-status.appspot.com/cq/henrik.smiding@intel.com/281963002/20001

[commit-bot: I haz the power, 2014-05-15 15:59:43]

Change committed as 14750