First pass at a blitter for R11 EAC alpha masks. This shaves 10ms off

First pass at a blitter for R11 EAC alpha masks. This shaves 10ms off
of the polygon gpu benchmark on the Nexus 7v2 (which is about 6.7% faster).

Committed: https://skia.googlesource.com/skia/+/ad1df153a16b7723d2d4afd13627919fc3ee87ce

[robertphillips, 2014-07-21 14:28:07]

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
File src/utils/SkTextureCompressor.cpp (right):

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:886: R11_EACBlitter::R11_EACBlitter(int width,
int height, void *latcBuffer)
Where is 0x7FFE coming from?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:902: (y-1 != fBufferedRuns[fNextRun-1].fY))) {
// comment ?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:907: while ((row + fNextRun) < y) {
// comment?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:933: 
left justify # lines ?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:976: 
Would making these two inlined functions kill performance?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompressor.h
File src/utils/SkTextureCompressor.h (right):

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:46: 
// This class ... ?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:51: 
/// -> // ?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:53: virtual void blitH(int x, int y, int width)
SK_OVERRIDE {
Is this allowed in the SkBlitter contract or a TODO ?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:64: virtual void blitV(int x, int y, int height,
SkAlpha alpha) SK_OVERRIDE {
Again - can we do without this?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:69: virtual void blitRect(int x, int y, int
width, int height) SK_OVERRIDE {
Here too?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:78: SkAlpha leftAlpha, SkAlpha rightAlpha)
SK_OVERRIDE {
?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:83: virtual void blitMask(const SkMask&, const
SkIRect& clip) SK_OVERRIDE {
?

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:129: 
order is supposed to be member variables then methods.

[krajcevski, 2014-07-21 17:35:27]

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
File src/utils/SkTextureCompressor.cpp (right):

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:886: R11_EACBlitter::R11_EACBlitter(int width,
int height, void *latcBuffer)
On 2014/07/21 14:28:06, robertphillips wrote:
> Where is 0x7FFE coming from?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:902: (y-1 != fBufferedRuns[fNextRun-1].fY))) {
On 2014/07/21 14:28:06, robertphillips wrote:
> // comment ?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:907: while ((row + fNextRun) < y) {
On 2014/07/21 14:28:06, robertphillips wrote:
> // comment?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:933: 
On 2014/07/21 14:28:06, robertphillips wrote:
> left justify # lines ?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.cpp:976: 
On 2014/07/21 14:28:06, robertphillips wrote:
> Would making these two inlined functions kill performance?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompressor.h
File src/utils/SkTextureCompressor.h (right):

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:46: 
On 2014/07/21 14:28:06, robertphillips wrote:
> // This class ... ?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:51: 
On 2014/07/21 14:28:06, robertphillips wrote:
> /// -> // ?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:53: virtual void blitH(int x, int y, int width)
SK_OVERRIDE {
On 2014/07/21 14:28:06, robertphillips wrote:
> Is this allowed in the SkBlitter contract or a TODO ?

All of the not implemented functions are more of a TODO. I think that they are
all required by the contract, but it really depends how you're going to use the
blitter. I.e. if this blitter gets used for a RGB bitmap, then it probably
should fail because this function isn't implemented.

I've added a comment to each fail to reflect this.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:64: virtual void blitV(int x, int y, int height,
SkAlpha alpha) SK_OVERRIDE {
On 2014/07/21 14:28:06, robertphillips wrote:
> Again - can we do without this?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:69: virtual void blitRect(int x, int y, int
width, int height) SK_OVERRIDE {
On 2014/07/21 14:28:06, robertphillips wrote:
> Here too?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:78: SkAlpha leftAlpha, SkAlpha rightAlpha)
SK_OVERRIDE {
On 2014/07/21 14:28:06, robertphillips wrote:
> ?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:83: virtual void blitMask(const SkMask&, const
SkIRect& clip) SK_OVERRIDE {
On 2014/07/21 14:28:06, robertphillips wrote:
> ?

Done.

https://codereview.chromium.org/406693002/diff/1/src/utils/SkTextureCompresso...
src/utils/SkTextureCompressor.h:129: 
On 2014/07/21 14:28:06, robertphillips wrote:
> order is supposed to be member variables then methods.

Done.

[krajcevski, 2014-07-21 17:38:00]

Forgot to upload patch set.

[robertphillips, 2014-07-21 18:08:25]

lgtm + some additional suggestions

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor.cpp (right):

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.cpp:808: 
// comment ?

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.cpp:942: // with zeros. We do this by producing a
single RLE that says
that we _have_ ?

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.cpp:1015: //                   that we're
operating on
curAlphai -> curAlphaColumn ?

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor.h (right):

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.h:179: // is large enough to store a compressed
image of size fWidth*fHeight.
add space after '*' ?

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.h:198: uint64_t *getBlock(int x, int y) const {
this->getBlockOffset ?

[krajcevski, 2014-07-21 18:24:20]

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor.cpp (right):

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.cpp:808: 
On 2014/07/21 18:08:25, robertphillips wrote:
> // comment ?

Done.

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.cpp:942: // with zeros. We do this by producing a
single RLE that says
On 2014/07/21 18:08:25, robertphillips wrote:
> that we _have_ ?

Done.

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.cpp:1015: //                   that we're
operating on
On 2014/07/21 18:08:25, robertphillips wrote:
> curAlphai -> curAlphaColumn ?

Done.

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor.h (right):

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.h:179: // is large enough to store a compressed
image of size fWidth*fHeight.
On 2014/07/21 18:08:25, robertphillips wrote:
> add space after '*' ?

Done.

https://codereview.chromium.org/406693002/diff/40001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.h:198: uint64_t *getBlock(int x, int y) const {
On 2014/07/21 18:08:25, robertphillips wrote:
> this->getBlockOffset ?

Done.

[krajcevski, 2014-07-21 18:25:35]

The CQ bit was checked by krajcevski@google.com

[commit-bot: I haz the power, 2014-07-21 18:26:18]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/krajcevski@google.com/406693002/80001

[krajcevski, 2014-07-21 18:32:45]

The CQ bit was checked by krajcevski@google.com

[commit-bot: I haz the power, 2014-07-21 18:32:57]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/krajcevski@google.com/406693002/100001

[commit-bot: I haz the power, 2014-07-21 18:44:42]

Change committed as ad1df153a16b7723d2d4afd13627919fc3ee87ce