Add color space xform support to SkJpegCodec (includes F16!)

Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed: https://skia.googlesource.com/skia/+/73d55332e2846dd05e9efdaa2f017bcc3872884b
Committed: https://skia.googlesource.com/skia/+/50ce1f28ffede3fa3e38d330d4114ee52b387848

[msarett, 2016-07-21 21:25:58]

Description was changed from

==========
Add color space xform support to SkJpegCodec (includes F16!)

BUG=skia:
==========

to

==========
Add color space xform support to SkJpegCodec (includes F16!)

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[msarett, 2016-07-22 13:59:52]

Patchset #1 (id:1) has been deleted

[msarett, 2016-07-22 15:27:51]

Patchset #1 (id:20001) has been deleted

[msarett, 2016-07-22 15:29:40]

Patchset #1 (id:40001) has been deleted

[msarett, 2016-07-22 15:38:48]

Description was changed from

==========
Add color space xform support to SkJpegCodec (includes F16!)

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
Add color space xform support to SkJpegCodec (includes F16!)


Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[msarett, 2016-07-22 15:38:57]

Description was changed from

==========
Add color space xform support to SkJpegCodec (includes F16!)


Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[msarett, 2016-07-22 15:39:45]

msarett@google.com changed reviewers:
+ brianosman@google.com, herb@google.com, mtklein@google.com, reed@google.com

[msarett, 2016-07-22 15:39:46]

https://codereview.chromium.org/2174493002/diff/60001/bench/ColorCodecBench.cpp
File bench/ColorCodecBench.cpp (left):

https://codereview.chromium.org/2174493002/diff/60001/bench/ColorCodecBench.c...
bench/ColorCodecBench.cpp:160: fSrcInfo =
codec->getInfo().makeColorType(kRGBA_8888_SkColorType);
This just moved with some minor modifications...  Now we can set the fDstSpace
on the fDstInfo.

https://codereview.chromium.org/2174493002/diff/60001/dm/DMSrcSink.cpp
File dm/DMSrcSink.cpp (left):

https://codereview.chromium.org/2174493002/diff/60001/dm/DMSrcSink.cpp#oldcod...
dm/DMSrcSink.cpp:870: SkImageInfo info =
codec->getInfo().makeColorType(fColorType);
Once again just moved this code block into a more logical order... now that we
want to put the dstSpace on the decodeInfo.

https://codereview.chromium.org/2174493002/diff/60001/src/codec/SkJpegCodec.cpp
File src/codec/SkJpegCodec.cpp (left):

https://codereview.chromium.org/2174493002/diff/60001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:402: // We will enable decodes to gray even if the
image is color because this is
This was a dumb feature that nobody uses and currently isn't being tested.

https://codereview.chromium.org/2174493002/diff/60001/src/codec/SkJpegCodec.cpp
File src/codec/SkJpegCodec.cpp (right):

https://codereview.chromium.org/2174493002/diff/60001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:487: JSAMPLE* decodeDst = (JSAMPLE*) dst;
Gladly taking suggestions to make this function more readable and/or efficient.

If we swizzle, it will never be "in place" (since it will probably scale and/or
subset).

We can color xform "in place", unless we must convert to F16 (in which case we
have a fColorXformSrcRow).

Probably the most common/important case will be:
(1) No swizzler
(2) Yes color xform (may be "in place" or not)

https://codereview.chromium.org/2174493002/diff/60001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:518: switch (dstInfo.colorType()) {
It'd be nice if we could call a function ptr here, but we need all the state on
the color xform object...

[msarett, 2016-07-25 21:04:12]

Rebasing this on other CLs that I think ought to land first.
I actually think this makes things simpler.

[mtklein, 2016-07-27 13:51:06]

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.cpp
File src/codec/SkJpegCodec.cpp (right):

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:387: fDecoderMgr->dinfo()->out_color_space =
JCS_EXT_RGBA;
// Our color transformation code requires RGBA order inputs, but it'll swizzle
to BGRA for us.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:418: // Disable F16 in legacy mode.
Yeah, the new standard line is

// F16 makes no sense without a color space.
return false;

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:480: return (kRGBA_F16_SkColorType ==
dstInfo.colorType()) ||
?

Is this, we know we need to end up in F16 by the time we're done, and the only
way to get there is by calling the colorspace transform code, even if the
transform itself will be an identity transform?

(... pipeline ...)

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:487: JSAMPLE* decodeDst = (JSAMPLE*) dst;
Some comments about what aliases what, when, and why it's safe would be nice
here.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:509: if (lines != 1) {
// i.e. == 0 ?

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:514: fSwizzler->swizzle(swizzleDst, decodeDst);
... oh hey, this would be a good pipeline stage too ...

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:571: std::move(sk_ref_sp(dstInfo.colorSpace())));
Aren't these moves redundant?

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:683: if (needsColorXform) {
Hmm.  I feel like I've read this before.  Think this chunk of logic wants to be
its own method called here and in onGetPixels()?

https://codereview.chromium.org/2174493002/diff/80001/src/opts/SkColorXform_o...
File src/opts/SkColorXform_opts.h (right):

https://codereview.chromium.org/2174493002/diff/80001/src/opts/SkColorXform_o...
src/opts/SkColorXform_opts.h:89: auto store_4 = [&dstReds, &dstGreens,
&dstBlues, &dst, &dstTables, &kRShift, &kGShift,
Probably clearer to close over logical constants by value (no &).

[msarett, 2016-07-27 14:37:18]

Patchset #3 (id:100001) has been deleted

[msarett, 2016-07-27 15:13:50]

On the RasterPipeline conversation:
I've been thinking more deeply about this, particularly now
that I've started work on PNG.
I'm enthusiastic about it from a design perspective, but I
don't feel confident that the performance is good
enough to hit a few goals.
(1) Color correct *all* images (not just the 10% with ICC
    profiles)
    RasterPipeline is about 50% slower on my desktop.
    This is significant given that I'm hoping to add
    a very minimal cost to decode time (to convince that
    it's ok to correct all images).
(2) Color correct images on Android
    Chrome currently has this disabled because QCMS is
    "too slow" on Android.  Given the problems with Arm v7
    code generation and Arm v8 (GCC) code generation, I
    don't think using RasterPipeline will allow us to claim
    that we are not "too slow".

I'm all for iterating to find a design that clearly deals
with the many possible xforms that we need.  But I'm not
convinced RasterPipeline is the answer yet.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.cpp
File src/codec/SkJpegCodec.cpp (right):

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:387: fDecoderMgr->dinfo()->out_color_space =
JCS_EXT_RGBA;
On 2016/07/27 13:51:06, mtklein wrote:
> // Our color transformation code requires RGBA order inputs, but it'll swizzle
> to BGRA for us.

Done.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:418: // Disable F16 in legacy mode.
On 2016/07/27 13:51:06, mtklein wrote:
> Yeah, the new standard line is
> 
> // F16 makes no sense without a color space.
> return false;

Deleting this whole block.  It's unnecessary because SkColorSpaceXform::New()
will return nullptr and we will catch this elsewhere.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:480: return (kRGBA_F16_SkColorType ==
dstInfo.colorType()) ||
On 2016/07/27 13:51:06, mtklein wrote:
> ?
> 
> Is this, we know we need to end up in F16 by the time we're done, and the only
> way to get there is by calling the colorspace transform code, even if the
> transform itself will be an identity transform?
> 
> (... pipeline ...)

The transform will never be an identity transform because we need to convert to
linear gamma.

The gamut transform may be identity.  As an optimization, we should catch this
and skip the step.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:487: JSAMPLE* decodeDst = (JSAMPLE*) dst;
On 2016/07/27 13:51:05, mtklein wrote:
> Some comments about what aliases what, when, and why it's safe would be nice
> here.

SGTM

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:509: if (lines != 1) {
On 2016/07/27 13:51:06, mtklein wrote:
> // i.e. == 0 ?

Done.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:514: fSwizzler->swizzle(swizzleDst, decodeDst);
On 2016/07/27 13:51:05, mtklein wrote:
> ... oh hey, this would be a good pipeline stage too ...

The swizzler performs sampling and subsetting.  It's potentially possible but
would be a bit tricky.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:571: std::move(sk_ref_sp(dstInfo.colorSpace())));
On 2016/07/27 13:51:06, mtklein wrote:
> Aren't these moves redundant?

Because we ref them as parameters?  Sure, seems like they ought to be. 
Removing.

https://codereview.chromium.org/2174493002/diff/80001/src/codec/SkJpegCodec.c...
src/codec/SkJpegCodec.cpp:683: if (needsColorXform) {
On 2016/07/27 13:51:06, mtklein wrote:
> Hmm.  I feel like I've read this before.  Think this chunk of logic wants to
be
> its own method called here and in onGetPixels()?

Done.

https://codereview.chromium.org/2174493002/diff/80001/src/opts/SkColorXform_o...
File src/opts/SkColorXform_opts.h (right):

https://codereview.chromium.org/2174493002/diff/80001/src/opts/SkColorXform_o...
src/opts/SkColorXform_opts.h:89: auto store_4 = [&dstReds, &dstGreens,
&dstBlues, &dst, &dstTables, &kRShift, &kGShift,
On 2016/07/27 13:51:06, mtklein wrote:
> Probably clearer to close over logical constants by value (no &).

Done.

[msarett, 2016-07-27 21:41:09]

Moving setjmp call to the appropriate function.

[msarett, 2016-07-28 16:27:53]

Fixed return value bug in readRows().

[mtklein, 2016-07-28 19:45:26]

https://codereview.chromium.org/2174493002/diff/160001/src/codec/SkJpegCodec.h
File src/codec/SkJpegCodec.h (right):

https://codereview.chromium.org/2174493002/diff/160001/src/codec/SkJpegCodec....
src/codec/SkJpegCodec.h:132: uint32_t*                         
fColorXformSrcRow;
Do we make sure the uint32_t* is 4 byte aligned into fStorage?

https://codereview.chromium.org/2174493002/diff/160001/src/core/SkColorSpaceX...
File src/core/SkColorSpaceXform.cpp (right):

https://codereview.chromium.org/2174493002/diff/160001/src/core/SkColorSpaceX...
src/core/SkColorSpaceXform.cpp:709: fStorage.reset(storageBytes);
I thought this was going on the stack?

[msarett, 2016-07-28 20:07:50]

Patchset #7 (id:200001) has been deleted

[msarett, 2016-07-28 20:09:46]

https://codereview.chromium.org/2174493002/diff/160001/src/codec/SkJpegCodec.h
File src/codec/SkJpegCodec.h (right):

https://codereview.chromium.org/2174493002/diff/160001/src/codec/SkJpegCodec....
src/codec/SkJpegCodec.h:132: uint32_t*                         
fColorXformSrcRow;
On 2016/07/28 19:45:26, mtklein wrote:
> Do we make sure the uint32_t* is 4 byte aligned into fStorage?

Should be fine because we always decode to RGBA (multiple of 4) when we do a
color xform.  But this result was more by luck than on purpose.  I've added an
SkASSERT.

https://codereview.chromium.org/2174493002/diff/160001/src/core/SkColorSpaceX...
File src/core/SkColorSpaceXform.cpp (right):

https://codereview.chromium.org/2174493002/diff/160001/src/core/SkColorSpaceX...
src/core/SkColorSpaceXform.cpp:709: fStorage.reset(storageBytes);
On 2016/07/28 19:45:26, mtklein wrote:
> I thought this was going on the stack?

Sorry this really wanted a rebase.  Should be fixed now.

[mtklein, 2016-07-28 20:10:34]

lgtm

[msarett, 2016-07-28 20:13:49]

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

[commit-bot: I haz the power, 2016-07-28 20:13:51]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[msarett, 2016-07-28 20:14:07]

Description was changed from

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

[commit-bot: I haz the power, 2016-07-28 20:54:06]

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

[commit-bot: I haz the power, 2016-07-28 20:54:07]

Dry run: Try jobs failed on following builders:
  Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Debug-Trybot on master.client.skia
(JOB_FAILED,
http://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX2...)

[msarett, 2016-07-28 21:13:03]

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

[commit-bot: I haz the power, 2016-07-28 21:13:15]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-28 21:14:45]

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

[commit-bot: I haz the power, 2016-07-28 21:14:46]

Dry run: Try jobs failed on following builders:
  skia_presubmit-Trybot on master.client.skia.fyi (JOB_FAILED,
http://build.chromium.org/p/client.skia.fyi/builders/skia_presubmit-Trybot/bu...)

[msarett, 2016-07-28 21:20:26]

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

[commit-bot: I haz the power, 2016-07-28 21:20:28]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-28 22:04:09]

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

[commit-bot: I haz the power, 2016-07-28 22:04:10]

Dry run: This issue passed the CQ dry run.

[msarett, 2016-07-28 22:05:10]

The CQ bit was checked by msarett@google.com

[msarett, 2016-07-28 22:05:11]

The patchset sent to the CQ was uploaded after l-g-t-m from mtklein@google.com
Link to the patchset: https://codereview.chromium.org/2174493002/#ps260001
(title: "Rebase")

[commit-bot: I haz the power, 2016-07-28 22:05:21]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-28 22:06:19]

Description was changed from

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
==========

to

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed:
https://skia.googlesource.com/skia/+/73d55332e2846dd05e9efdaa2f017bcc3872884b
==========

[commit-bot: I haz the power, 2016-07-28 22:06:20]

Committed patchset #9 (id:260001) as
https://skia.googlesource.com/skia/+/73d55332e2846dd05e9efdaa2f017bcc3872884b

[msarett, 2016-07-29 00:11:04]

A revert of this CL (patchset #9 id:260001) has been created in
https://codereview.chromium.org/2195523002/ by msarett@google.com.

The reason for reverting is: Breaking MSAN.

[mtklein, 2016-07-29 01:26:12]

https://codereview.chromium.org/2174493002/diff/260001/bench/ColorCodecBench.cpp
File bench/ColorCodecBench.cpp (right):

https://codereview.chromium.org/2174493002/diff/260001/bench/ColorCodecBench....
bench/ColorCodecBench.cpp:176: } else if (FLAGS_qcms) {
This is used without checking defined(SK_TEST_QCMS).

[msarett, 2016-07-29 12:25:34]

Description was changed from

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed:
https://skia.googlesource.com/skia/+/73d55332e2846dd05e9efdaa2f017bcc3872884b
==========

to

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed:
https://skia.googlesource.com/skia/+/73d55332e2846dd05e9efdaa2f017bcc3872884b
==========

[msarett, 2016-07-29 12:38:40]

Patchset #12 (id:320001) has been deleted

[msarett, 2016-07-29 12:38:48]

https://codereview.chromium.org/2174493002/diff/260001/bench/ColorCodecBench.cpp
File bench/ColorCodecBench.cpp (right):

https://codereview.chromium.org/2174493002/diff/260001/bench/ColorCodecBench....
bench/ColorCodecBench.cpp:176: } else if (FLAGS_qcms) {
On 2016/07/29 01:26:12, mtklein wrote:
> This is used without checking defined(SK_TEST_QCMS).

Ahh thanks.

[msarett, 2016-07-29 12:39:07]

The CQ bit was checked by msarett@google.com

[msarett, 2016-07-29 12:39:08]

The patchset sent to the CQ was uploaded after l-g-t-m from mtklein@google.com
Link to the patchset: https://codereview.chromium.org/2174493002/#ps300001
(title: "Fix MSAN suppression")

[commit-bot: I haz the power, 2016-07-29 12:39:12]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-29 13:23:35]

Description was changed from

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed:
https://skia.googlesource.com/skia/+/73d55332e2846dd05e9efdaa2f017bcc3872884b
==========

to

==========
Add color space xform support to SkJpegCodec (includes F16!)

Also changes SkColorXform to support:
RGBA->RGBA
RGBA->BGRA

Instead of:
RGBA->SkPMColor

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2174493002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Committed:
https://skia.googlesource.com/skia/+/73d55332e2846dd05e9efdaa2f017bcc3872884b
Committed:
https://skia.googlesource.com/skia/+/50ce1f28ffede3fa3e38d330d4114ee52b387848
==========

[commit-bot: I haz the power, 2016-07-29 13:23:36]

Committed patchset #11 (id:300001) as
https://skia.googlesource.com/skia/+/50ce1f28ffede3fa3e38d330d4114ee52b387848