lots of sRGB and F16 blits

lots of sRGB and F16 blits

- generalize F16 xfermode procs
- spriteblits for F16 and sRGB
- saveLayer now respects colortype and profiletype

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1686013002

Committed: https://skia.googlesource.com/skia/+/129ed1cd6d792f3f6cf563aefa9756fc6308289d

[reed1, 2016-02-10 19:34:38]

Description was changed from

==========
generalize xferu64 procs using proc4f

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
generalize xferu64 procs using proc4f

cloned from https://codereview.chromium.org/1686013002/

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[reed1, 2016-02-10 19:59:19]

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

[commit-bot: I haz the power, 2016-02-10 19:59:34]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685203002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685203002/20001

[commit-bot: I haz the power, 2016-02-10 20:27:22]

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

[commit-bot: I haz the power, 2016-02-10 20:27:22]

Dry run: This issue passed the CQ dry run.

[reed1, 2016-02-19 19:28:16]

reed@google.com changed reviewers:
+ fmalita@chromium.org, herb@google.com, mtklein@google.com

[reed1, 2016-02-19 19:28:16]

ptal

[reed1, 2016-02-19 19:28:22]

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

[commit-bot: I haz the power, 2016-02-19 19:28:27]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685203002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685203002/40001

[reed1, 2016-02-19 19:39:47]

This one was fun...

To test the full half-float blend modes, I change the GM to use saveLayer to
show the effects...

This made me fix saveLayer to *not* downgrade to 32bit in the layer

This made me fix sprite-blitters to know about half-floats (and sRGB src)

[commit-bot: I haz the power, 2016-02-19 20:06:17]

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

[commit-bot: I haz the power, 2016-02-19 20:06:18]

Dry run: This issue passed the CQ dry run.

[reed1, 2016-02-19 20:14:30]

reed@google.com changed reviewers:
+ brianosman@google.com

[f(malita), 2016-02-19 20:38:43]

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
File src/core/SkSpriteBlitterU64.cpp (right):

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
src/core/SkSpriteBlitterU64.cpp:30: (SkNx_cast<float>(Sk4b::Load(&addr[i])) *
Sk4f(1.0f/255)).store(span[i].fVec);
Is there a benefit to loading Sk4f(1/255) outside the loop, or is Sk4f
compiler-friendly enough to optimize such const loads anyway?

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
src/core/SkSpriteBlitterU64.cpp:146: SkPM4f*                 fBuffer;
SkAutoTMalloc?

[mtklein, 2016-02-19 20:41:17]

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
File src/core/SkSpriteBlitterU64.cpp (right):

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
src/core/SkSpriteBlitterU64.cpp:30: (SkNx_cast<float>(Sk4b::Load(&addr[i])) *
Sk4f(1.0f/255)).store(span[i].fVec);
On 2016/02/19 20:38:43, f(malita) wrote:
> Is there a benefit to loading Sk4f(1/255) outside the loop, or is Sk4f
> compiler-friendly enough to optimize such const loads anyway?

It's very hoistable.  I'd only move it out of the loop if you'd otherwise think
it's more readable there.

[f(malita), 2016-02-19 20:58:31]

On 2016/02/19 20:41:17, mtklein wrote:
>
https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
> File src/core/SkSpriteBlitterU64.cpp (right):
> 
>
https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
> src/core/SkSpriteBlitterU64.cpp:30: (SkNx_cast<float>(Sk4b::Load(&addr[i])) *
> Sk4f(1.0f/255)).store(span[i].fVec);
> On 2016/02/19 20:38:43, f(malita) wrote:
> > Is there a benefit to loading Sk4f(1/255) outside the loop, or is Sk4f
> > compiler-friendly enough to optimize such const loads anyway?
> 
> It's very hoistable.  I'd only move it out of the loop if you'd otherwise
think
> it's more readable there.

Awesome, good to know.  While on the subject, how about
constant-folding/identity ops?

Should I try to avoid stuff like

foo4f * Sk4f(bar)

where bar is compile-time == 1.0f?

[mtklein, 2016-02-19 21:15:18]

On 2016/02/19 20:58:31, f(malita) wrote:
> On 2016/02/19 20:41:17, mtklein wrote:
> >
>
https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
> > File src/core/SkSpriteBlitterU64.cpp (right):
> > 
> >
>
https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
> > src/core/SkSpriteBlitterU64.cpp:30: (SkNx_cast<float>(Sk4b::Load(&addr[i]))
*
> > Sk4f(1.0f/255)).store(span[i].fVec);
> > On 2016/02/19 20:38:43, f(malita) wrote:
> > > Is there a benefit to loading Sk4f(1/255) outside the loop, or is Sk4f
> > > compiler-friendly enough to optimize such const loads anyway?
> > 
> > It's very hoistable.  I'd only move it out of the loop if you'd otherwise
> think
> > it's more readable there.
> 
> Awesome, good to know.  While on the subject, how about
> constant-folding/identity ops?
> 
> Should I try to avoid stuff like
> 
> foo4f * Sk4f(bar)
> 
> where bar is compile-time == 1.0f?

As far as I can tell, math follow the same rules as for ordinary floats, which
are surprisingly subtle.
It's always worth hacking up a little test case and disassembling it to take a
look.

I think in both cases the multiply must happen, as multiplying by 1 is not an
identity operation in floats.  (Nor is adding 0.)

[herb_g, 2016-02-19 21:36:46]

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
File src/core/SkSpriteBlitterU64.cpp (right):

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
src/core/SkSpriteBlitterU64.cpp:6: * found in the LICENSE file.
I think this should be:
/*
 * Copyright 2016 Google Inc.
 *
 * Use of this source code is governed by a BSD-style license that can be
 * found in the LICENSE file.
 */

[reed1, 2016-02-20 19:37:17]

ptal

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
File src/core/SkSpriteBlitterU64.cpp (right):

https://codereview.chromium.org/1685203002/diff/40001/src/core/SkSpriteBlitte...
src/core/SkSpriteBlitterU64.cpp:6: * found in the LICENSE file.
On 2016/02/19 21:36:46, herb_g wrote:
> I think this should be:
> /*
>  * Copyright 2016 Google Inc.
>  *
>  * Use of this source code is governed by a BSD-style license that can be
>  * found in the LICENSE file.
>  */

Done.

[reed1, 2016-02-20 19:37:46]

Description was changed from

==========
generalize xferu64 procs using proc4f

cloned from https://codereview.chromium.org/1686013002/

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
lots of sRGB and F16 blits

cloned from https://codereview.chromium.org/1686013002/

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[reed1, 2016-02-20 19:37:54]

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

[commit-bot: I haz the power, 2016-02-20 19:38:06]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685203002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685203002/60001

[reed1, 2016-02-20 19:38:44]

Description was changed from

==========
lots of sRGB and F16 blits

cloned from https://codereview.chromium.org/1686013002/

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
lots of sRGB and F16 blits

- generalize F16 xfermode procs
- spriteblits for F16 and sRGB
- saveLayer now respects colortype and profiletype

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

[commit-bot: I haz the power, 2016-02-20 20:01:21]

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

[commit-bot: I haz the power, 2016-02-20 20:01:22]

Dry run: This issue passed the CQ dry run.

[reed1, 2016-02-20 20:30:12]

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

[commit-bot: I haz the power, 2016-02-20 20:30:21]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685203002/80001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685203002/80001

[commit-bot: I haz the power, 2016-02-20 20:53:10]

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

[commit-bot: I haz the power, 2016-02-20 20:53:10]

Dry run: This issue passed the CQ dry run.

[reed1, 2016-02-21 03:32:12]

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

[commit-bot: I haz the power, 2016-02-21 03:32:57]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685203002/100001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685203002/100001

[reed1, 2016-02-21 03:40:08]

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

[commit-bot: I haz the power, 2016-02-21 03:40:13]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685203002/120001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685203002/120001

[commit-bot: I haz the power, 2016-02-21 04:03:40]

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

[commit-bot: I haz the power, 2016-02-21 04:03:41]

Dry run: This issue passed the CQ dry run.

[f(malita), 2016-02-22 13:58:18]

lgtm

https://codereview.chromium.org/1685203002/diff/120001/src/core/SkSpriteBlitt...
File src/core/SkSpriteBlitter4f.cpp (right):

https://codereview.chromium.org/1685203002/diff/120001/src/core/SkSpriteBlitt...
src/core/SkSpriteBlitter4f.cpp:17: fBuffer = new SkPM4f[src.width()];
Nit: SkAutoTMalloc instead of manual management?

[reed1, 2016-02-22 14:17:22]

https://codereview.chromium.org/1685203002/diff/120001/src/core/SkSpriteBlitt...
File src/core/SkSpriteBlitter4f.cpp (right):

https://codereview.chromium.org/1685203002/diff/120001/src/core/SkSpriteBlitt...
src/core/SkSpriteBlitter4f.cpp:17: fBuffer = new SkPM4f[src.width()];
On 2016/02/22 13:58:18, f(malita) wrote:
> Nit: SkAutoTMalloc instead of manual management?

Done.

[reed1, 2016-02-22 14:17:32]

The CQ bit was checked by reed@google.com

[reed1, 2016-02-22 14:17:32]

The patchset sent to the CQ was uploaded after l-g-t-m from fmalita@chromium.org
Link to the patchset: https://codereview.chromium.org/1685203002/#ps140001
(title: "use SkAutoTMalloc")

[commit-bot: I haz the power, 2016-02-22 14:17:38]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685203002/140001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685203002/140001

[commit-bot: I haz the power, 2016-02-22 14:42:34]

Description was changed from

==========
lots of sRGB and F16 blits

- generalize F16 xfermode procs
- spriteblits for F16 and sRGB
- saveLayer now respects colortype and profiletype

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...
==========

to

==========
lots of sRGB and F16 blits

- generalize F16 xfermode procs
- spriteblits for F16 and sRGB
- saveLayer now respects colortype and profiletype

BUG=skia:
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Committed:
https://skia.googlesource.com/skia/+/129ed1cd6d792f3f6cf563aefa9756fc6308289d
==========

[commit-bot: I haz the power, 2016-02-22 14:42:35]

Committed patchset #8 (id:140001) as
https://skia.googlesource.com/skia/+/129ed1cd6d792f3f6cf563aefa9756fc6308289d