Add preliminary ASTC encoder

Add preliminary ASTC encoder

Committed: https://skia.googlesource.com/skia/+/b2ef181ed614d1a5772367a8cd7ab77aa015ae29

[robertphillips, 2014-07-24 17:47:40]

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor.h (right):

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.h:24: 
Change kLast_Format too ?

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor_ASTC.cpp (right):

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:135: uint8_t get_alpha(const uint8_t
*src, int rowBytes, int x, int y) {
SkASSERT the x & y assumptions here ?

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:142: uint8_t indices[5][6];
Any reason you have the outer loop variable as 'j' instead of 'i' ?
Would 'x' and 'y' be meaningful here ?

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:145: const int idx = j*6+i;
weightTot, alphaTot ?

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:180: // The rest of the block contains
the 30 index values from before, which
typo

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:191: // need to have their bits reversed
I don't know if this can land without a bit twiddling version of this.

[krajcevski, 2014-07-24 19:59:08]

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor.h (right):

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor.h:24: 
On 2014/07/24 17:47:39, robertphillips wrote:
> Change kLast_Format too ?

Done.

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor_ASTC.cpp (right):

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:135: uint8_t get_alpha(const uint8_t
*src, int rowBytes, int x, int y) {
On 2014/07/24 17:47:39, robertphillips wrote:
> SkASSERT the x & y assumptions here ?

Done.

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:142: uint8_t indices[5][6];
On 2014/07/24 17:47:39, robertphillips wrote:
> Any reason you have the outer loop variable as 'j' instead of 'i' ?
> Would 'x' and 'y' be meaningful here ?

Actually, we can just loop from 0 to 30, since we're processing the indices in
order...

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:145: const int idx = j*6+i;
On 2014/07/24 17:47:39, robertphillips wrote:
> weightTot, alphaTot ?

Done.

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:180: // The rest of the block contains
the 30 index values from before, which
On 2014/07/24 17:47:39, robertphillips wrote:
> typo

Done.

https://codereview.chromium.org/417933005/diff/20001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:191: // need to have their bits reversed
On 2014/07/24 17:47:39, robertphillips wrote:
> I don't know if this can land without a bit twiddling version of this.

Yea, that's kind of ugly. I changed it so that we streamline the process a bit
better. Doing a reverse of a 3-bit value just takes a swap, and we can reverse
all three-bit values in parallel with a couple of xors.

[robertphillips, 2014-07-25 13:06:58]

lgtm + two questions

https://codereview.chromium.org/417933005/diff/60001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor_ASTC.cpp (right):

https://codereview.chromium.org/417933005/diff/60001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:184: const int weight =
k6x5To12x12Table[idx][w*3];
Can we move the x & y lookups inside the "if (weight > 0)" test?

https://codereview.chromium.org/417933005/diff/60001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:255: uint8_t** dstPtr = &dst;
j,i -> y,x ?

[krajcevski, 2014-07-25 14:20:14]

https://codereview.chromium.org/417933005/diff/60001/src/utils/SkTextureCompr...
File src/utils/SkTextureCompressor_ASTC.cpp (right):

https://codereview.chromium.org/417933005/diff/60001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:184: const int weight =
k6x5To12x12Table[idx][w*3];
On 2014/07/25 13:06:58, robertphillips wrote:
> Can we move the x & y lookups inside the "if (weight > 0)" test?

Done.

https://codereview.chromium.org/417933005/diff/60001/src/utils/SkTextureCompr...
src/utils/SkTextureCompressor_ASTC.cpp:255: uint8_t** dstPtr = &dst;
On 2014/07/25 13:06:58, robertphillips wrote:
> j,i -> y,x ?

Done.

[krajcevski, 2014-07-25 14:20:17]

The CQ bit was checked by krajcevski@google.com

[commit-bot: I haz the power, 2014-07-25 14:20:29]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/krajcevski@google.com/417933005/80001

[commit-bot: I haz the power, 2014-07-25 14:33:06]

Change committed as b2ef181ed614d1a5772367a8cd7ab77aa015ae29