Add MSA optimized I422ToARGBRow_MSA and I422ToRGBARow_MSA functions

Add MSA optimized I422ToARGBRow_MSA and I422ToRGBARow_MSA functions

R=fbarchard@google.com
BUG=libyuv:634
Performance Gains :- (vs C vectorized)

I422ToARGBRow_MSA     : ~1.6x
I422ToRGBARow_MSA     : ~1.6x

I422ToARGBRow_Any_MSA : ~1.58x
I422ToRGBARow_Any_MSA : ~1.6x


Performance Gains :- (vs C non-vectorized)

I422ToARGBRow_MSA     : ~7x
I422ToRGBARow_MSA     : ~7x

I422ToARGBRow_Any_MSA : ~6.9x
I422ToRGBARow_Any_MSA : ~6.8x

Regarding performance measurement, We have created standalone tests which pass in row's data from a 1920x1080 filled buffer to both the C and MSA functions. And such N iterations are executed to get more accurate timings of C vs MSA.

Committed: https://chromium.googlesource.com/libyuv/libyuv/+/f5d5bd88d660232038fe06ed735fe95d2b9f9b61

[manojkumar.bhosale, 2016-10-20 13:08:00]

Description was changed from

==========
Add MSA optimized I422ToARGBRow_MSA and I422ToRGBARow_MSA functions

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

to

==========
Add MSA optimized I422ToARGBRow_MSA and I422ToRGBARow_MSA functions

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


Performance Gains :- (vs C vectorized)

I422ToARGBRow_MSA     : ~1.6x
I422ToRGBARow_MSA     : ~1.6x

I422ToARGBRow_Any_MSA : ~1.58x
I422ToRGBARow_Any_MSA : ~1.6x


Performance Gains :- (vs C non-vectorized)

I422ToARGBRow_MSA     : ~7x
I422ToRGBARow_MSA     : ~7x

I422ToARGBRow_Any_MSA : ~6.9x
I422ToRGBARow_Any_MSA : ~6.8x

Regarding performance measurement, We have created standalone tests which pass
in row's data from a 1920x1080 filled buffer to both the C and MSA functions.
And such N iterations are executed to get more accurate timings of C vs MSA.
==========

[manojkumar.bhosale, 2016-10-20 13:08:00]

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

[fbarchard1, 2016-10-24 17:36:32]

lgtm

just minor suggestions

https://codereview.chromium.org/2430313005/diff/1/include/libyuv/macros_msa.h
File include/libyuv/macros_msa.h (right):

https://codereview.chromium.org/2430313005/diff/1/include/libyuv/macros_msa.h...
include/libyuv/macros_msa.h:25: \
remove blank line

https://codereview.chromium.org/2430313005/diff/1/source/row_msa.cc
File source/row_msa.cc (right):

https://codereview.chromium.org/2430313005/diff/1/source/row_msa.cc#newcode22
source/row_msa.cc:22: #define I422TORGB(in0, in1, in2, ub, vr, ug, vg,          
    \
Could use a brief comment to describe what this macro does.  What are in/out? 
How many are done?

https://codereview.chromium.org/2430313005/diff/1/source/row_msa.cc#newcode186
source/row_msa.cc:186: data_y = LD(src_y);
Consider a macro to read I422, as there are many YUV formats converted to
ARGB... e.g. NV12, with the read here the only difference.

https://codereview.chromium.org/2430313005/diff/1/source/row_msa.cc#newcode191
source/row_msa.cc:191: src2 = (v16u8) __msa_insert_w(zero, 0, data_v);
No way to load 4 or 8 bytes directly into an MSA register?
On Intel there is movd and movq, or you can pinsrw to insert a word.
On Arm the float/double registers are the same as Neon, or you can use an array
notation.

[fbarchard1, 2016-10-24 22:33:28]

On 2016/10/24 17:36:32, fbarchard1 wrote:
> lgtm
> 
> just minor suggestions
> 
> https://codereview.chromium.org/2430313005/diff/1/include/libyuv/macros_msa.h
> File include/libyuv/macros_msa.h (right):
> 
>
https://codereview.chromium.org/2430313005/diff/1/include/libyuv/macros_msa.h...
> include/libyuv/macros_msa.h:25: \
> remove blank line
> 
> https://codereview.chromium.org/2430313005/diff/1/source/row_msa.cc
> File source/row_msa.cc (right):
> 
> https://codereview.chromium.org/2430313005/diff/1/source/row_msa.cc#newcode22
> source/row_msa.cc:22: #define I422TORGB(in0, in1, in2, ub, vr, ug, vg,        
 
>     \
> Could use a brief comment to describe what this macro does.  What are in/out? 
> How many are done?
> 
> https://codereview.chromium.org/2430313005/diff/1/source/row_msa.cc#newcode186
> source/row_msa.cc:186: data_y = LD(src_y);
> Consider a macro to read I422, as there are many YUV formats converted to
> ARGB... e.g. NV12, with the read here the only difference.
> 
> https://codereview.chromium.org/2430313005/diff/1/source/row_msa.cc#newcode191
> source/row_msa.cc:191: src2 = (v16u8) __msa_insert_w(zero, 0, data_v);
> No way to load 4 or 8 bytes directly into an MSA register?
> On Intel there is movd and movq, or you can pinsrw to insert a word.
> On Arm the float/double registers are the same as Neon, or you can use an
array
> notation.

I'll go ahead and land this, but follow up next CL related to YUV to RGB, with
comments etc.

[fbarchard1, 2016-10-24 22:37:11]

Description was changed from

==========
Add MSA optimized I422ToARGBRow_MSA and I422ToRGBARow_MSA functions

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


Performance Gains :- (vs C vectorized)

I422ToARGBRow_MSA     : ~1.6x
I422ToRGBARow_MSA     : ~1.6x

I422ToARGBRow_Any_MSA : ~1.58x
I422ToRGBARow_Any_MSA : ~1.6x


Performance Gains :- (vs C non-vectorized)

I422ToARGBRow_MSA     : ~7x
I422ToRGBARow_MSA     : ~7x

I422ToARGBRow_Any_MSA : ~6.9x
I422ToRGBARow_Any_MSA : ~6.8x

Regarding performance measurement, We have created standalone tests which pass
in row's data from a 1920x1080 filled buffer to both the C and MSA functions.
And such N iterations are executed to get more accurate timings of C vs MSA.
==========

to

==========
Add MSA optimized I422ToARGBRow_MSA and I422ToRGBARow_MSA functions

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


Performance Gains :- (vs C vectorized)

I422ToARGBRow_MSA     : ~1.6x
I422ToRGBARow_MSA     : ~1.6x

I422ToARGBRow_Any_MSA : ~1.58x
I422ToRGBARow_Any_MSA : ~1.6x


Performance Gains :- (vs C non-vectorized)

I422ToARGBRow_MSA     : ~7x
I422ToRGBARow_MSA     : ~7x

I422ToARGBRow_Any_MSA : ~6.9x
I422ToRGBARow_Any_MSA : ~6.8x

Regarding performance measurement, We have created standalone tests which pass
in row's data from a 1920x1080 filled buffer to both the C and MSA functions.
And such N iterations are executed to get more accurate timings of C vs MSA.

Committed:
https://chromium.googlesource.com/libyuv/libyuv/+/f5d5bd88d660232038fe06ed735...
==========

[fbarchard1, 2016-10-24 22:37:12]

Committed patchset #1 (id:1) manually as
f5d5bd88d660232038fe06ed735fe95d2b9f9b61 (presubmit successful).