Small optimization for convolution shader: only apply the bounds check in

Small optimization for convolution shader: only apply the bounds check in direction of convolution, not both. This requires generating different versions of the shader depending on direction.

R=robertphillips@google.com

Committed: https://code.google.com/p/skia/source/detail?r=10417

[Stephen White, 2013-07-29 17:52:58]

Robert: PTAL. Thanks!

[robertphillips, 2013-07-29 18:02:38]

Mainly concerned about the [4] sized parameters

https://codereview.chromium.org/20789003/diff/5001/src/gpu/effects/GrConvolut...
File src/gpu/effects/GrConvolutionEffect.cpp (right):

https://codereview.chromium.org/20789003/diff/5001/src/gpu/effects/GrConvolut...
src/gpu/effects/GrConvolutionEffect.cpp:36: bool useBounds() const { return
fUseBounds; }
const?

https://codereview.chromium.org/20789003/diff/5001/src/gpu/effects/GrConvolut...
src/gpu/effects/GrConvolutionEffect.cpp:166: bool useBounds,
should this be [2]?

https://codereview.chromium.org/20789003/diff/5001/src/gpu/effects/GrConvolut...
src/gpu/effects/GrConvolutionEffect.cpp:182: bool useBounds,
should this be [2]?

[Stephen White, 2013-07-29 18:08:07]

Thanks for the review!

On 2013/07/29 18:02:38, robertphillips wrote:
> Mainly concerned about the [4] sized parameters
> 
>
https://codereview.chromium.org/20789003/diff/5001/src/gpu/effects/GrConvolut...
> File src/gpu/effects/GrConvolutionEffect.cpp (right):
> 
>
https://codereview.chromium.org/20789003/diff/5001/src/gpu/effects/GrConvolut...
> src/gpu/effects/GrConvolutionEffect.cpp:36: bool useBounds() const { return
> fUseBounds; }
> const?

I assume you mean direction()?  Fixed.

>
https://codereview.chromium.org/20789003/diff/5001/src/gpu/effects/GrConvolut...
> src/gpu/effects/GrConvolutionEffect.cpp:166: bool useBounds,
> should this be [2]?
> 
>
https://codereview.chromium.org/20789003/diff/5001/src/gpu/effects/GrConvolut...
> src/gpu/effects/GrConvolutionEffect.cpp:182: bool useBounds,
> should this be [2]?

Yeah, fixed Patch Set 4 (noticed and uploaded just after I sent the mail, doh..
:) )

[robertphillips, 2013-07-29 18:27:15]

lgtm

[Stephen White, 2013-07-29 18:45:49]

Committed patchset #5 manually as r10417 (presubmit successful).