Add MSA optimized ARGB Attenuate/RGB565/Shuffle/Shader/Gray/Sepia row functions

Add MSA optimized ARGB Attenuate/RGB565/Shuffle/Shader/Gray/Sepia row functions

R=fbarchard@google.com
BUG=libyuv:634
Performance Gain (vs C vectorized)
ARGBAttenuateRow_MSA          - ~1.1x
ARGBAttenuateRow_Any_MSA      - ~1.1x
ARGBToRGB565DitherRow_MSA     - ~6.4x
ARGBToRGB565DitherRow_Any_MSA - ~6.2x
ARGBShuffleRow_MSA            - ~5.1x
ARGBShuffleRow_Any_MSA        - ~1.9x
ARGBShadeRow_MSA              - ~1.1x
ARGBGrayRow_MSA               - ~2.6x
ARGBSepiaRow_MSA              - ~11.6x

Performance Gain (vs C non-vectorized)
ARGBAttenuateRow_MSA          - ~2.46x
ARGBAttenuateRow_Any_MSA      - ~2.45x
ARGBToRGB565DitherRow_MSA     - ~9.4x
ARGBToRGB565DitherRow_Any_MSA - ~12.5x
ARGBShuffleRow_MSA            - ~5.2x
ARGBShuffleRow_Any_MSA        - ~1.9x
ARGBShadeRow_MSA              - ~4.3x
ARGBGrayRow_MSA               - ~10.5x
ARGBSepiaRow_MSA              - ~12.2x

Committed: https://chromium.googlesource.com/libyuv/libyuv/+/a899dea25167749697ecd97de57e07db9d85a841

[manojkumar.bhosale, 2016-12-07 09:19:12]

Description was changed from

==========
Add MSA optimized ARGB Attenuate/RGB565/Shuffle/Shader/Gray/Sepia row functions

R=fbarchard@google.com
BUG=libyuv:634
==========

to

==========
Add MSA optimized ARGB Attenuate/RGB565/Shuffle/Shader/Gray/Sepia row functions

R=fbarchard@google.com
BUG=libyuv:634

Performance Gain (vs C vectorized)
ARGBAttenuateRow_MSA          - ~1.1x
ARGBAttenuateRow_Any_MSA      - ~1.1x
ARGBToRGB565DitherRow_MSA     - ~6.4x
ARGBToRGB565DitherRow_Any_MSA - ~6.2x
ARGBShuffleRow_MSA            - ~5.1x
ARGBShuffleRow_Any_MSA        - ~1.9x
ARGBShadeRow_MSA              - ~1.1x
ARGBGrayRow_MSA               - ~2.6x
ARGBSepiaRow_MSA              - ~11.6x

Performance Gain (vs C non-vectorized)
ARGBAttenuateRow_MSA          - ~2.46x
ARGBAttenuateRow_Any_MSA      - ~2.45x
ARGBToRGB565DitherRow_MSA     - ~9.4x
ARGBToRGB565DitherRow_Any_MSA - ~12.5x
ARGBShuffleRow_MSA            - ~5.2x
ARGBShuffleRow_Any_MSA        - ~1.9x
ARGBShadeRow_MSA              - ~4.3x
ARGBGrayRow_MSA               - ~10.5x
ARGBSepiaRow_MSA              - ~12.2x
==========

[manojkumar.bhosale, 2016-12-07 09:19:12]

manojkumar.bhosale@imgtec.com changed reviewers:
+ kjellander@chromium.org

[manojkumar.bhosale, 2016-12-07 09:19:38]

Updated with reviewers, cc list and performance gain numbers

[kjellander_chromium, 2016-12-07 09:20:02]

kjellander@chromium.org changed reviewers:
- kjellander@chromium.org

[kjellander_chromium, 2016-12-07 09:20:51]

On 2016/12/07 09:19:38, manojkumar.bhosale wrote:
> Updated with reviewers, cc list and performance gain numbers

No need to include me as reviewer on CLs like this. I don't know much of the
implementations in libyuv, I'm mostly owner in order to do infrastructure
changes.

[manojkumar.bhosale, 2016-12-07 09:24:33]

On 2016/12/07 09:20:51, kjellander_chromium wrote:
> On 2016/12/07 09:19:38, manojkumar.bhosale wrote:
> > Updated with reviewers, cc list and performance gain numbers
> 
> No need to include me as reviewer on CLs like this. I don't know much of the
> implementations in libyuv, I'm mostly owner in order to do infrastructure
> changes.

OK. Thanks.

[fbarchard1, 2016-12-12 20:09:03]

lgtm

https://codereview.chromium.org/2559693002/diff/1/source/row_msa.cc
File source/row_msa.cc (right):

https://codereview.chromium.org/2559693002/diff/1/source/row_msa.cc#newcode1050
source/row_msa.cc:1050: src0 = (v16u8)__msa_ld_b((v16i8*)src_argb, 0);
Can this function be further optimized?  Suggest a TODO for now, but when used,
this function tends to be used alot for converting ARGB to a form that can be
alpha blended with premultiplied values on each frame.

https://codereview.chromium.org/2559693002/diff/1/source/row_msa.cc#newcode1218
source/row_msa.cc:1218: v16u8 const_0x4B0F = (v16u8)__msa_fill_h(0x4B0F);
note these constants may need to be passed in as parameters in future.

[manojkumar.bhosale, 2016-12-15 06:34:52]

Description was changed from

==========
Add MSA optimized ARGB Attenuate/RGB565/Shuffle/Shader/Gray/Sepia row functions

R=fbarchard@google.com
BUG=libyuv:634

Performance Gain (vs C vectorized)
ARGBAttenuateRow_MSA          - ~1.1x
ARGBAttenuateRow_Any_MSA      - ~1.1x
ARGBToRGB565DitherRow_MSA     - ~6.4x
ARGBToRGB565DitherRow_Any_MSA - ~6.2x
ARGBShuffleRow_MSA            - ~5.1x
ARGBShuffleRow_Any_MSA        - ~1.9x
ARGBShadeRow_MSA              - ~1.1x
ARGBGrayRow_MSA               - ~2.6x
ARGBSepiaRow_MSA              - ~11.6x

Performance Gain (vs C non-vectorized)
ARGBAttenuateRow_MSA          - ~2.46x
ARGBAttenuateRow_Any_MSA      - ~2.45x
ARGBToRGB565DitherRow_MSA     - ~9.4x
ARGBToRGB565DitherRow_Any_MSA - ~12.5x
ARGBShuffleRow_MSA            - ~5.2x
ARGBShuffleRow_Any_MSA        - ~1.9x
ARGBShadeRow_MSA              - ~4.3x
ARGBGrayRow_MSA               - ~10.5x
ARGBSepiaRow_MSA              - ~12.2x
==========

to

==========
Add MSA optimized ARGB Attenuate/RGB565/Shuffle/Shader/Gray/Sepia row functions

R=fbarchard@google.com
BUG=libyuv:634

Performance Gain (vs C vectorized)
ARGBAttenuateRow_MSA          - ~1.1x
ARGBAttenuateRow_Any_MSA      - ~1.1x
ARGBToRGB565DitherRow_MSA     - ~6.4x
ARGBToRGB565DitherRow_Any_MSA - ~6.2x
ARGBShuffleRow_MSA            - ~5.1x
ARGBShuffleRow_Any_MSA        - ~1.9x
ARGBShadeRow_MSA              - ~1.1x
ARGBGrayRow_MSA               - ~2.6x
ARGBSepiaRow_MSA              - ~11.6x

Performance Gain (vs C non-vectorized)
ARGBAttenuateRow_MSA          - ~2.46x
ARGBAttenuateRow_Any_MSA      - ~2.45x
ARGBToRGB565DitherRow_MSA     - ~9.4x
ARGBToRGB565DitherRow_Any_MSA - ~12.5x
ARGBShuffleRow_MSA            - ~5.2x
ARGBShuffleRow_Any_MSA        - ~1.9x
ARGBShadeRow_MSA              - ~4.3x
ARGBGrayRow_MSA               - ~10.5x
ARGBSepiaRow_MSA              - ~12.2x

Committed:
https://chromium.googlesource.com/libyuv/libyuv/+/a899dea25167749697ecd97de57...
==========

[manojkumar.bhosale, 2016-12-15 06:34:57]

Committed patchset #1 (id:1) manually as
a899dea25167749697ecd97de57e07db9d85a841 (presubmit successful).

[manojkumar.bhosale, 2016-12-15 06:35:21]

https://codereview.chromium.org/2559693002/diff/1/source/row_msa.cc
File source/row_msa.cc (right):

https://codereview.chromium.org/2559693002/diff/1/source/row_msa.cc#newcode1050
source/row_msa.cc:1050: src0 = (v16u8)__msa_ld_b((v16i8*)src_argb, 0);
On 2016/12/12 20:09:03, fbarchard1 wrote:
> Can this function be further optimized?  Suggest a TODO for now, but when
used,
> this function tends to be used alot for converting ARGB to a form that can be
> alpha blended with premultiplied values on each frame.

Acknowledged.

https://codereview.chromium.org/2559693002/diff/1/source/row_msa.cc#newcode1218
source/row_msa.cc:1218: v16u8 const_0x4B0F = (v16u8)__msa_fill_h(0x4B0F);
On 2016/12/12 20:09:03, fbarchard1 wrote:
> note these constants may need to be passed in as parameters in future.

Acknowledged.