Improve performance of highQualityFilter

Improve performance of highQualityFilter

Move loop invariant caculation out of loop. This gives about 15% speedup
for bitmap_BGRA_8888_A_scale_rotate_bicubic on my desktop i7-3770.

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/d3f6e893df53fb0cfeefd01a8a49a77c72d742bd

[qiankun, 2014-08-26 10:12:10]

qiankun.miao@intel.com changed reviewers:
+ humper@google.com, mtklein@google.com

[qiankun, 2014-08-26 10:12:10]

PTAL.

[mtklein, 2014-08-26 13:17:52]

https://codereview.chromium.org/502953004/diff/1/src/core/SkBitmapFilter.cpp
File src/core/SkBitmapFilter.cpp (right):

https://codereview.chromium.org/502953004/diff/1/src/core/SkBitmapFilter.cpp#...
src/core/SkBitmapFilter.cpp:31: SkScalar* xWeights = SkNEW_ARRAY(SkScalar,
maxX);
I'd prefer

SkAutoTMalloc<SkScalar> xWeights(maxX);

from there all the code should be the same, except no need to SkDELETE_ARRAY the
thing.

https://codereview.chromium.org/502953004/diff/1/src/core/SkBitmapFilter.cpp#...
src/core/SkBitmapFilter.cpp:48: for (int srcX = x0; srcX < x1 ; srcX++) {
Add a note?
// Looking these up once instead of each loop is a ~15% speedup.

[reed1, 2014-08-26 14:17:14]

reed@google.com changed reviewers:
+ reed@google.com

[reed1, 2014-08-26 14:17:14]

15% sounds good

https://codereview.chromium.org/502953004/diff/1/src/core/SkBitmapFilter.cpp
File src/core/SkBitmapFilter.cpp (right):

https://codereview.chromium.org/502953004/diff/1/src/core/SkBitmapFilter.cpp#...
src/core/SkBitmapFilter.cpp:31: SkScalar* xWeights = SkNEW_ARRAY(SkScalar,
maxX);
On 2014/08/26 13:17:52, mtklein wrote:
> I'd prefer
> 
> SkAutoTMalloc<SkScalar> xWeights(maxX);
> 
> from there all the code should be the same, except no need to SkDELETE_ARRAY
the
> thing.

Or SkAutoTArray, which also gives you (debugging only) range checks for []

[qiankun, 2014-08-26 14:26:50]

https://codereview.chromium.org/502953004/diff/1/src/core/SkBitmapFilter.cpp
File src/core/SkBitmapFilter.cpp (right):

https://codereview.chromium.org/502953004/diff/1/src/core/SkBitmapFilter.cpp#...
src/core/SkBitmapFilter.cpp:31: SkScalar* xWeights = SkNEW_ARRAY(SkScalar,
maxX);
On 2014/08/26 13:17:52, mtklein wrote:
> I'd prefer
> 
> SkAutoTMalloc<SkScalar> xWeights(maxX);
> 
> from there all the code should be the same, except no need to SkDELETE_ARRAY
the
> thing.

Done.

https://codereview.chromium.org/502953004/diff/1/src/core/SkBitmapFilter.cpp#...
src/core/SkBitmapFilter.cpp:48: for (int srcX = x0; srcX < x1 ; srcX++) {
On 2014/08/26 13:17:52, mtklein wrote:
> Add a note?
> // Looking these up once instead of each loop is a ~15% speedup.

Done.

[mtklein, 2014-08-26 14:38:34]

lgtm

[qiankun, 2014-08-27 01:58:53]

The CQ bit was checked by qiankun.miao@intel.com

[commit-bot: I haz the power, 2014-08-27 01:59:21]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/qiankun.miao@intel.com/502953004/20001

[commit-bot: I haz the power, 2014-08-27 02:06:48]

Committed patchset #2 (20001) as d3f6e893df53fb0cfeefd01a8a49a77c72d742bd