div255(x) as ((x+128)*257)>>16 with SSE

div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*257)>>16 part simple.
This seems to speed up every transfermode that uses div255(),
in the 7-25% range.

It even appears to obviate the need for approxMulDiv255() on SSE.
I'm not sure about NEON yet, so I'll keep approxMulDiv255() for now.

Should be no pixels change:
https://gold.skia.org/search2?issue=1452903004&unt=true&query=source_type%3Dgm&master=false

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed: https://skia.googlesource.com/skia/+/cbf4fba43933302a846872e4c5ce8f1adb8b325e

[mtklein, 2015-11-17 21:37:12]

Description was changed from

==========
div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*256)>>16 part simple.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*257)>>16 part simple.
This seems to speed up every transfermode that uses div255().

It even appears to obviate the need for approxMulDiv255() on SSE.
I'm still working on speeding up div255() on NEON, so I'll want
to keep approxMulDiv255() around until I can get rid of it everywhere
so pixels stay consistent.

Should be no pixels change.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[mtklein, 2015-11-17 21:46:43]

The CQ bit was checked by mtklein@google.com to run a CQ dry run

[commit-bot: I haz the power, 2015-11-17 21:46:48]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1452903004/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1452903004/40001

[mtklein, 2015-11-17 21:51:10]

Description was changed from

==========
div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*257)>>16 part simple.
This seems to speed up every transfermode that uses div255().

It even appears to obviate the need for approxMulDiv255() on SSE.
I'm still working on speeding up div255() on NEON, so I'll want
to keep approxMulDiv255() around until I can get rid of it everywhere
so pixels stay consistent.

Should be no pixels change.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*257)>>16 part simple.
This seems to speed up every transfermode that uses div255(),
in the 7-25% range.

It even appears to obviate the need for approxMulDiv255() on SSE.
I'm still working on speeding up div255() on NEON, so I'll want
to keep approxMulDiv255() around until I can get rid of it everywhere
so pixels stay consistent.

Should be no pixels change.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[mtklein, 2015-11-17 21:54:10]

Description was changed from

==========
div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*257)>>16 part simple.
This seems to speed up every transfermode that uses div255(),
in the 7-25% range.

It even appears to obviate the need for approxMulDiv255() on SSE.
I'm still working on speeding up div255() on NEON, so I'll want
to keep approxMulDiv255() around until I can get rid of it everywhere
so pixels stay consistent.

Should be no pixels change.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*257)>>16 part simple.
This seems to speed up every transfermode that uses div255(),
in the 7-25% range.

It even appears to obviate the need for approxMulDiv255() on SSE.
I'm not sure about NEON yet, so I'll keep approxMulDiv255() for now.

Should be no pixels change.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[mtklein, 2015-11-17 21:54:47]

Description was changed from

==========
div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*257)>>16 part simple.
This seems to speed up every transfermode that uses div255(),
in the 7-25% range.

It even appears to obviate the need for approxMulDiv255() on SSE.
I'm not sure about NEON yet, so I'll keep approxMulDiv255() for now.

Should be no pixels change.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
div255(x) as ((x+128)*257)>>16 with SSE

_mm_mulhi_epu16 makes the (...*257)>>16 part simple.
This seems to speed up every transfermode that uses div255(),
in the 7-25% range.

It even appears to obviate the need for approxMulDiv255() on SSE.
I'm not sure about NEON yet, so I'll keep approxMulDiv255() for now.

Should be no pixels change:
https://gold.skia.org/search2?issue=1452903004&unt=true&query=source_type%3Dg...

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[mtklein_C, 2015-11-17 21:56:07]

mtklein@chromium.org changed reviewers:
+ reed@google.com
- mtklein@google.com

[mtklein_C, 2015-11-17 21:56:07]

This is a follow up from that unit test CL from this morning.

[commit-bot: I haz the power, 2015-11-17 21:58:39]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-11-17 21:58:39]

Dry run: This issue passed the CQ dry run.

[reed1, 2015-11-17 22:12:21]

wow wow wow
lgtm

[mtklein, 2015-11-17 22:19:05]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2015-11-17 22:19:15]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1452903004/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1452903004/40001

[commit-bot: I haz the power, 2015-11-17 22:19:56]

Committed patchset #3 (id:40001) as
https://skia.googlesource.com/skia/+/cbf4fba43933302a846872e4c5ce8f1adb8b325e