Add MIPS SIMD Arch (MSA) optimized ARGBMirrorRow function

Add MIPS SIMD Arch (MSA) optimized ARGBMirrorRow function

This patch adds MSA optimized ARGBMirrorRow function in libYUV project.

Performance gain ~3x

R=fbarchard@google.com
BUG=libyuv:634
Committed: https://chromium.googlesource.com/libyuv/libyuv/+/618149084eaaa2b572417aaed0419834f1d91449

[manojkumar.bhosale, 2016-09-26 16:47:27]

Description was changed from

==========
Add MIPS SIMD Arch (MSA) optimized ARGBMirrorRow function

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

to

==========
Add MIPS SIMD Arch (MSA) optimized ARGBMirrorRow function

This patch adds MSA optimized ARGBMirrorRow function in libYUV project.

Performance gain ~3x

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

[manojkumar.bhosale, 2016-09-26 16:47:27]

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

[fbarchard1, 2016-09-26 18:04:34]

lgtm

Looks good but I'd suggest less unrolling and avoid macros.

https://codereview.chromium.org/2368313003/diff/1/include/libyuv/macros_msa.h
File include/libyuv/macros_msa.h (right):

https://codereview.chromium.org/2368313003/diff/1/include/libyuv/macros_msa.h...
include/libyuv/macros_msa.h:79: out0 = (RTYPE) __msa_vshf_w((v4i32) mask0,
(v4i32) in1, (v4i32) in0);  \
style nit - I'd prefer we dont end up with macros for every MSA instruction,
unrolled in various ways.  Consider using __msa_vshf_w directly in row_msa.cc

https://codereview.chromium.org/2368313003/diff/1/source/row_msa.cc
File source/row_msa.cc (right):

https://codereview.chromium.org/2368313003/diff/1/source/row_msa.cc#newcode42
source/row_msa.cc:42: int count;
prefer int x for horizontal counts for consistency with row_common.cc

for (x = 0; x < width; x += 2) {

https://codereview.chromium.org/2368313003/diff/1/source/row_msa.cc#newcode51
source/row_msa.cc:51: VSHF_W4_UB(src0, src0, src1, src1, src2, src2, src3, src3,
consider less unrolling.  (measure performance)

For this function on other CPUs just 1 or 2 shuffles is enough to achieve most
of the performance without excessive unrolling which hurts performance on odd
widths.
Rule of thumb is do 2 shuffles
LD_UB2
VSHF_W2_UB
ST_UB2

8 pixels at a time, 32 bytes.

[fbarchard1, 2016-09-26 23:28:03]

Description was changed from

==========
Add MIPS SIMD Arch (MSA) optimized ARGBMirrorRow function

This patch adds MSA optimized ARGBMirrorRow function in libYUV project.

Performance gain ~3x

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

to

==========
Add MIPS SIMD Arch (MSA) optimized ARGBMirrorRow function

This patch adds MSA optimized ARGBMirrorRow function in libYUV project.

Performance gain ~3x

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

Committed:
https://chromium.googlesource.com/libyuv/libyuv/+/618149084eaaa2b572417aaed04...
==========

[fbarchard1, 2016-09-26 23:28:05]

Committed patchset #1 (id:1) manually as
618149084eaaa2b572417aaed0419834f1d91449 (presubmit successful).

[fbarchard1, 2016-09-27 00:33:26]

ARGBMirrorRow_MSA changed to match MirrorRow_MSA with different shuffler
constants.
Will follow up with removing some unused macros.