Revert "Start of rewrite of GrFragmentProcessor optimizations."

This reverts commit 85eb4226a4cd8c10a0e3f3ba2f3a60efbb2dd61b.

Reason for revert: test failures on Windows, e.g. https://chromium-swarm.appspot.com/task?id=33f9527484414110&refresh=10

Original change's description:
> Start of rewrite of GrFragmentProcessor optimizations.
>
> This adds a replacement for computeInvariantOutput buts does not use it yet. The replacement allows for three types of optimizations:
>
> * known input color -> known output color for GrFP elimination
> * tracking of whether all color processors modulate their input for the "tweak alpha" optimziation
> * opaqueness tracking
>
> This loses some of the generality of computInvariantOutput. It does not track the known output status of individual color components (other than opaque alpha). It does not track whether GrFragmentProcessors read their input color. It doesn't allow a processor that will receive non-constant output to advertise that it produces a constant output. These could probably be added back in the unlikely case that they prove valuable.
>
> Unlike computeInvariantOutput the optimizations are decided at instantiation time and constant colors are expressed as GrColor4f rather than GrColor.
>
> Change-Id: I684d3f9050693dde2d28154fa695e049ed8cf61a
> Reviewed-on: https://skia-review.googlesource.com/7481
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
>

TBR=egdaniel@google.com,bsalomon@google.com,brianosman@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: I2390df257456013fa74137cb5d7b5a93820c291e
Reviewed-on: https://skia-review.googlesource.com/7652
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>

Start of rewrite of GrFragmentProcessor optimizations.

This adds a replacement for computeInvariantOutput buts does not use it yet. The replacement allows for three types of optimizations:

* known input color -> known output color for GrFP elimination
* tracking of whether all color processors modulate their input for the "tweak alpha" optimziation
* opaqueness tracking

This loses some of the generality of computInvariantOutput. It does not track the known output status of individual color components (other than opaque alpha). It does not track whether GrFragmentProcessors read their input color. It doesn't allow a processor that will receive non-constant output to advertise that it produces a constant output. These could probably be added back in the unlikely case that they prove valuable.

Unlike computeInvariantOutput the optimizations are decided at instantiation time and constant colors are expressed as GrColor4f rather than GrColor.

Change-Id: I684d3f9050693dde2d28154fa695e049ed8cf61a
Reviewed-on: https://skia-review.googlesource.com/7481
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>

Revert "Revert "SkUserConfig.h: remove out-of-date comment on SK_SFNTLY_SUBSETTER""

This reverts commit 156a749eb15d18539a58436d767e9401de413eb3.

Reason for revert: Reverting this makes it much easier to test the merge_into_android.py script. I also stopped the Android autoroller for now.

Original change's description:
> Revert "SkUserConfig.h: remove out-of-date comment on SK_SFNTLY_SUBSETTER"
>
> This reverts commit 68fc549e840d1efc6285c301da1bf52c04e57e0a.
>
> Reason for revert: Breaking the Android roll.
>
> (The change is fine, but the script for merging did not consider that we might edit this file. We can reland this after we update the script.)
>
> Original change's description:
> > SkUserConfig.h: remove out-of-date comment on SK_SFNTLY_SUBSETTER
> >
> > Change-Id: I6894fe43de071999e4923e1515a951b73e2ba0b1
> > Reviewed-on: https://skia-review.googlesource.com/7619
> > Reviewed-by: Ben Wagner <bungeman@google.com>
> > Commit-Queue: Hal Canary <halcanary@google.com>
> >
>
> TBR=halcanary@google.com,bungeman@google.com,reed@google.com
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
>
> Change-Id: Ib1a2c9faa79d1186705d87c5e17f20fe49b82bb1
> Reviewed-on: https://skia-review.googlesource.com/7641
> Reviewed-by: Leon Scroggins <scroggo@google.com>
> Commit-Queue: Leon Scroggins <scroggo@google.com>
>

TBR=mtklein@chromium.org,halcanary@google.com,bungeman@google.com,scroggo@google.com,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: Ifa42e4fb48bbe9dac40c57cf09cf9586ca86c6df
Reviewed-on: https://skia-review.googlesource.com/7649
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>

Revert "Consolidate read/write-Pixels testing code"

This reverts commit 0bd4a23ef40d74ef051eb7f457c133d5febc377d.

Reason for revert: What?!?

Original change's description:
> Consolidate read/write-Pixels testing code
>
> Change-Id: I853f8f747ed0040333473fbc722cabac84e6ac83
> Reviewed-on: https://skia-review.googlesource.com/7560
> Commit-Queue: Robert Phillips <robertphillips@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
>

TBR=robertphillips@google.com,brianosman@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: I9cb049a4f86350f849d00839a55b55fb07e861b0
Reviewed-on: https://skia-review.googlesource.com/7648
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>

remove SK_SUPPORT_LEGACY_GETCLIPBOUNDS

BUG=skia:6166

Change-Id: I5c8035ee172e14a7a769baff45392a29014d09cb
Reviewed-on: https://skia-review.googlesource.com/7645
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>

Consolidate read/write-Pixels testing code

Change-Id: I853f8f747ed0040333473fbc722cabac84e6ac83
Reviewed-on: https://skia-review.googlesource.com/7560
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>

Revert "SkUserConfig.h: remove out-of-date comment on SK_SFNTLY_SUBSETTER"

This reverts commit 68fc549e840d1efc6285c301da1bf52c04e57e0a.

Reason for revert: Breaking the Android roll.

(The change is fine, but the script for merging did not consider that we might edit this file. We can reland this after we update the script.)

Original change's description:
> SkUserConfig.h: remove out-of-date comment on SK_SFNTLY_SUBSETTER
>
> Change-Id: I6894fe43de071999e4923e1515a951b73e2ba0b1
> Reviewed-on: https://skia-review.googlesource.com/7619
> Reviewed-by: Ben Wagner <bungeman@google.com>
> Commit-Queue: Hal Canary <halcanary@google.com>
>

TBR=halcanary@google.com,bungeman@google.com,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: Ib1a2c9faa79d1186705d87c5e17f20fe49b82bb1
Reviewed-on: https://skia-review.googlesource.com/7641
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>

Delete SkAutoTime from jpeg encoder

We use nanobench instead to measure encode performance.

BUG=skia:

Change-Id: I4f0b43b7e09f99967889e941cb6947433033006a
Reviewed-on: https://skia-review.googlesource.com/7640
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>

Keep valgrind happen by purging caches.

Change-Id: Ic387edb9b5d71110c10c8a40a362ed2853d140f5
Reviewed-on: https://skia-review.googlesource.com/7638
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Herb Derby <herb@google.com>

Implement getFamilyName for stream fonts on Linux.

When SkFontMgr_fontconfig and SkFontMgr_FontConfigInterface create a
typeface from data they do not store the default font name and
getFamilyName will return the empty string. All of the code to handle
this properly now exists, it just needs to be hooked up.

BUG=skia:1508

Change-Id: I75f2a598a5451babb4a9ceb5e9a9e9d3daa41d60
Reviewed-on: https://skia-review.googlesource.com/7506
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Ben Wagner <bungeman@google.com>

Simplify SkShader::ContextRec::fLocalMatrix handling

Minor cleanup.

R=robertphillips@google.com,reed@google.com

Change-Id: I1a27af61e3e3bb5e895c9a3265da6b0408234165
Reviewed-on: https://skia-review.googlesource.com/7632
Commit-Queue: Florin Malita <fmalita@chromium.org>
Reviewed-by: Robert Phillips <robertphillips@google.com>

Really use vpmaddwd in hsw::convolve_vertical().

No pixel diffs.
Performance on 8888 looks like an overall win.

Before:
    micros  bench
    222.41  bitmap_scale_filter_64_256
     40.06  bitmap_scale_filter_256_64
      8.17  bitmap_scale_filter_90_10
     10.32  bitmap_scale_filter_90_30
     22.50  bitmap_scale_filter_90_80
      1.80  bitmap_scale_filter_90_90
     57.51  bitmap_scale_filter_80_90
     41.99  bitmap_scale_filter_30_90
     31.51  bitmap_scale_filter_10_90

After:
    micros  bench
    193.60  bitmap_scale_filter_64_256
     46.26  bitmap_scale_filter_256_64
      7.81  bitmap_scale_filter_90_10
      9.99  bitmap_scale_filter_90_30
     22.05  bitmap_scale_filter_90_80
      1.96  bitmap_scale_filter_90_90
     52.07  bitmap_scale_filter_80_90
     37.73  bitmap_scale_filter_30_90
     27.63  bitmap_scale_filter_10_90

Change-Id: I2f29366b0fd503176c5af4d825fa524e632da21b
Reviewed-on: https://skia-review.googlesource.com/7630
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Graceful degredation for SkOpts_hsw.

__AVX2__ will not be defined if you omit -mavx2.  Android does this
intentionally for x86 builds.  (No mobile CPU supports AVX2 AFAIK.)

This should fix the Android roll.

Change-Id: Ib94c862641abc11fbb46863afc53bcc049f362ad
Reviewed-on: https://skia-review.googlesource.com/7633
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Matt Sarett <msarett@google.com>

Reland "remove clipbounds flag"

This reverts commit 2aa26abe6daad2fc933ef638992cce71cb65f458.

Reason for revert: CL this depends on has now landed.

Original change's description:
> Revert "remove clipbounds flag"
>
> This reverts commit 7e0489da25c578bd04ce74d03ad48198280ca3fd.
>
> Reason for revert: missed a headless
>
> Original change's description:
> > remove clipbounds flag
> >
> > BUG=skia:
> >
> > Change-Id: I55b3098abd5e566e9f8bf27988c33e531d2ea54d
> > Reviewed-on: https://skia-review.googlesource.com/7618
> > Reviewed-by: Mike Reed <reed@google.com>
> > Commit-Queue: Mike Reed <reed@google.com>
> >
>
> TBR=reed@google.com,reviews@skia.org
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=skia:
>
> Change-Id: Id1310372499905dddcca4b9454f6a81d192a1e32
> Reviewed-on: https://skia-review.googlesource.com/7629
> Commit-Queue: Mike Reed <reed@google.com>
> Reviewed-by: Mike Reed <reed@google.com>
>

TBR=reviews@skia.org,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Change-Id: I593bdad8d654e272e772559aa0fccafcce8ff541
Reviewed-on: https://skia-review.googlesource.com/7636
Commit-Queue: Ben Wagner <benjaminwagner@google.com>
Reviewed-by: Ben Wagner <benjaminwagner@google.com>

Fix gradients gm

Some of my local changes slipped into the tree.

BUG=skia:

Change-Id: I6e452f12608db6b80c5887ccd3cac199cfbf9494
Reviewed-on: https://skia-review.googlesource.com/7625
Reviewed-by: Matt Sarett <msarett@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Matt Sarett <msarett@google.com>

Revert "remove clipbounds flag"

This reverts commit 7e0489da25c578bd04ce74d03ad48198280ca3fd.

Reason for revert: missed a headless

Original change's description:
> remove clipbounds flag
>
> BUG=skia:
>
> Change-Id: I55b3098abd5e566e9f8bf27988c33e531d2ea54d
> Reviewed-on: https://skia-review.googlesource.com/7618
> Reviewed-by: Mike Reed <reed@google.com>
> Commit-Queue: Mike Reed <reed@google.com>
>

TBR=reed@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Change-Id: Id1310372499905dddcca4b9454f6a81d192a1e32
Reviewed-on: https://skia-review.googlesource.com/7629
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>

Fix variable names in convolve_vertical().

These new names reflect the actual pixels stored in each register.

Change-Id: I8e626196cd8bcbef622e4fb87ac3566a79d3573a
Reviewed-on: https://skia-review.googlesource.com/7624
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Matt Sarett <msarett@google.com>

extend clipper for lines, for uniformity, and so curves can call itself with lines

BUG=skia:

Change-Id: Id8c1fba6fcd2496802d3d17afe3f5c91bf5dfc33
Reviewed-on: https://skia-review.googlesource.com/7621
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Mike Reed <reed@google.com>

SkUserConfig.h: remove out-of-date comment on SK_SFNTLY_SUBSETTER

Change-Id: I6894fe43de071999e4923e1515a951b73e2ba0b1
Reviewed-on: https://skia-review.googlesource.com/7619
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Hal Canary <halcanary@google.com>

exclude *hsw* source files on Android/iOS G3 build.

Like *sse2*, *avx*, etc, these are x86-specific and
even there must be compiled with particular flags.

Change-Id: Ibbc004b9ba5da6da70449177423a785d2c644846
Reviewed-on: https://skia-review.googlesource.com/7623
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Add SkICC::rawTransferFnData()

BUG=skia:

Change-Id: I912b044a0091a4d396c954d1ad1d84f5f8d50f56
Reviewed-on: https://skia-review.googlesource.com/7366
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>

remove clipbounds flag

BUG=skia:

Change-Id: I55b3098abd5e566e9f8bf27988c33e531d2ea54d
Reviewed-on: https://skia-review.googlesource.com/7618
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>

remove obsolete flag

Added by humper in 2013, it is not
referenced anywhere I can find.

R=reed@google.com

Change-Id: I35b3891b29b2fddb132647b5a1e337a0fc68e79e
Reviewed-on: https://skia-review.googlesource.com/7610
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>

SkOpts_hsw ODR paranoia

I'm warming back up to the idea of very careful use of SkOpts_hsw.
But if we're going to do that, we need a strict header discipline.
No header can be assumed to be safe without vetting, and most aren't.

Today there's only one function defined in SkOpts_hsw, so this CL mostly
rewrites that convolve_vertically() to use no headers beyond immintrin.h
and stdint.h, both safe.  It shared very little code with the others
anyway, so we're not losing anything by putting it directly into
SkOpts_hsw.cpp.  I have also streamlined the implementation considerably
to improve maintainability and readability.

Change-Id: Ia03daae660e54125a0d2e2988464cfc930349e80
Reviewed-on: https://skia-review.googlesource.com/7611
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Remove SkBlitter::getShaderContext()

Unused.  Also ComposeShaderContext::getShaderContext{A,B}.

R=reed@google.com,mtklein@google.com

Change-Id: I36a8346f0c7e842455774339f388c04920db0040
Reviewed-on: https://skia-review.googlesource.com/7613
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Florin Malita <fmalita@chromium.org>

Fix extension suffic for getting procs from GL_EXT_debug_marker on ES

Change-Id: I483fbf7c5f6d4489802def78fa855f70313e4a19
Reviewed-on: https://skia-review.googlesource.com/7602
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>

add test for big cubic -- currently disabled

BUG=skia:

Change-Id: Ief11a1e54d427b1241b47f3cfab5b44fb656443a
Reviewed-on: https://skia-review.googlesource.com/7607
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>

Move SkGammaColorFilter to tools, limit to sRGB

Similarly, limit GrGammaEffect to sRGB (and rename it).

BUG=skia:

Change-Id: I88feef11ab7040bca2fa4c2eed71923ded87a0d0
Reviewed-on: https://skia-review.googlesource.com/7375
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>

remove legacy clipbounds flag for android

BUG=skia:

Change-Id: Ie13c85bfbb5b1b173e8ac6a2fb0c9a23ea8ba7de
Reviewed-on: https://skia-review.googlesource.com/7583
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Mike Reed <reed@google.com>

Move GrDrawingManager pointer to GrSurfaceContext

Change-Id: Ic285b24b384bbf284cc680fe770433dd4d643833
Reviewed-on: https://skia-review.googlesource.com/7561
Reviewed-by: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>

Check that the upper bits are correct in the SkArenaAlloc footer.

TBR=mtklein@google.com

Change-Id: I8dc21778d1b8b24dd9d7b6919bb7e2bcfe090e82
Reviewed-on: https://skia-review.googlesource.com/7588
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Herb Derby <herb@google.com>

Fix clang warning about unused lambda capture.

This was found by a tip of tree build of clang.

Change-Id: I9aa3fecd47ae69350e8d58e796f960a991b7759e
Reviewed-on: https://skia-review.googlesource.com/7586
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Remove redundant caps field from GrProcessorTestData

Change-Id: Ic31338f08d1bd5b55ffab75d30c0f54436c8b917
Reviewed-on: https://skia-review.googlesource.com/7510
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>

Simplify Footers in SkArenaAlloc to 64-bit values using pointers.

TBR=mtklein@google.com

Change-Id: I72c6cf6857b2bb00f4259cc9c4de2d51d454e6ab
Reviewed-on: https://skia-review.googlesource.com/7582
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Herb Derby <herb@google.com>

Switch GrConfigConversionEffect over to taking GrTextureProxies

Change-Id: Ic8be773e210e1ac05dcb9aad6c89dcd63e9e4ba2
Reviewed-on: https://skia-review.googlesource.com/7521
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>

utils/win/SkTScopedComPtr<T>: Add C++11ish move semantics

Change-Id: I026ffa81d583fddc4a2c7238a176232ef98fcd89
Reviewed-on: https://skia-review.googlesource.com/7434
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Hal Canary <halcanary@google.com>

Use GrSurfaceContext::read/write-Pixels more

Change-Id: I04bcaf91aa7a43e0563e332c1fe2836d762a04d4
Reviewed-on: https://skia-review.googlesource.com/7520
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>

Add test for bad reuse of SkJpegCodec internal state

This is a test for the following fix:
https://skia-review.googlesource.com/c/7451/

b/34637813

Change-Id: I3fdb0d7d50537948039b75c3a22454e789ca810a
Reviewed-on: https://skia-review.googlesource.com/7454
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>

Move SkShadowTessellator to GrShadowTessellator

BUG=skia:6119

Change-Id: I9756b5aeaced1f21a65063470ccb013c7b856f28
Reviewed-on: https://skia-review.googlesource.com/7505
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>

Roll recipe DEPS

BUG=skia:

Change-Id: If10fe8b3691c8a71b52774f763a08120ef20dfd8
Reviewed-on: https://skia-review.googlesource.com/7502
Reviewed-by: Ravi Mistry <rmistry@google.com>
Commit-Queue: Eric Boren <borenet@google.com>

Add geometric implementation for ambient shadows

Original: https://skia-review.googlesource.com/7273

BUG=skia:6119

Change-Id: Ie7debd7727768f55dafad922a2b8b9fd3a638fda
Reviewed-on: https://skia-review.googlesource.com/7500
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>

Revert "Remove SkImage::MakeTextureFromPixmap"

This reverts commit 8ccbbb455c6ef194920c47e846b95b50f5017138.

Reason for revert: Chrome perf regression

BUG=chrome:684483

Original change's description:
> Remove SkImage::MakeTextureFromPixmap
>
> Another step towards reducing the number of texture upload paths.
>
> BUG=skia:
>
> Change-Id: Ica185e7334f52dc9ebf87e21fe6f60589ef87bb3
> Reviewed-on: https://skia-review.googlesource.com/7346
> Commit-Queue: Brian Osman <brianosman@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
>

TBR=bsalomon@google.com,robertphillips@google.com,brianosman@google.com,cblume@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
BUG=skia:

Change-Id: I8bf6970b764018f3ebcac8123e4d6a8ad05317ff
Reviewed-on: https://skia-review.googlesource.com/7484
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>

Revert "Add geometric implementation for ambient shadows"

This reverts commit fdb1bdf1aa4e7f14befcb3acc3f586e7734190ea.

Reason for revert: Breaking the roll (see e.g. https://codereview.chromium.org/2652833006)

Original change's description:
> Add geometric implementation for ambient shadows
>
> BUG=skia:6119
>
> Change-Id: I3140522f223c35fc059a33b593064897485dff7c
> Reviewed-on: https://skia-review.googlesource.com/7273
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Jim Van Verth <jvanverth@google.com>
>

TBR=jvanverth@google.com,bsalomon@google.com,robertphillips@google.com,reviews@skia.org,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:6119

Change-Id: I307f6084c40e5ba305c38d15b6b433465295553c
Reviewed-on: https://skia-review.googlesource.com/7480
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>

Add geometric implementation for ambient shadows

BUG=skia:6119

Change-Id: I3140522f223c35fc059a33b593064897485dff7c
Reviewed-on: https://skia-review.googlesource.com/7273
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>

Revert "Remove sentinal GLContext used in early days of Vulkan in our tools"

This reverts commit bfd97c59350c112ad85301eb6ff6005ae4fd9e8c.

Reason for revert: Windows bots crashing.

Original change's description:
> Remove sentinal GLContext used in early days of Vulkan in our tools
>
> I've ran this through multiple variations of dm and nanobench on Linux
> and no longer see any crash/hangs. I forget what the original repo case
> for the bug was, but I'm at least not seeing it now with updated drivers.
>
> BUG=skia:
>
> Change-Id: I6b7129a4c4d67938baa35d2e2c720cb078fc4c18
> Reviewed-on: https://skia-review.googlesource.com/7441
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Greg Daniel <egdaniel@google.com>
>

TBR=egdaniel@google.com,bsalomon@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Change-Id: I94a16b2b13d09182232061b0aebd0d58df96db8a
Reviewed-on: https://skia-review.googlesource.com/7453
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>

Allow GrSingleTextureEffect to take GrTextureProxies

Change-Id: I1dd441a5838f665c6815a5c629f5763f43f66e09
Reviewed-on: https://skia-review.googlesource.com/7429
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>

Use appropriate options in SkJpegCodec readRows()

b/34637813

Change-Id: Ibebac2ef8cfe004beb2272c266b580946cedcb08
Reviewed-on: https://skia-review.googlesource.com/7451
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>

Make the footer big to handle bug 684061.

R=mtklein@google.com

BUG=chromium:684061

Change-Id: I210ca93e00ec6fa5d166c0fde5cdcff0f60c26ca
Reviewed-on: https://skia-review.googlesource.com/7450
Commit-Queue: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Remove sentinal GLContext used in early days of Vulkan in our tools

I've ran this through multiple variations of dm and nanobench on Linux
and no longer see any crash/hangs. I forget what the original repo case
for the bug was, but I'm at least not seeing it now with updated drivers.

BUG=skia:

Change-Id: I6b7129a4c4d67938baa35d2e2c720cb078fc4c18
Reviewed-on: https://skia-review.googlesource.com/7441
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>

SkPixelInfo::CopyPixels() refactor part 0: only one memcpy() case

I envision this eventually looking like the following:
Fast Path 1: memcpy
Fast Path 2: Legacy RGBA/BGRA unpremul/premul
Fast Path 3: SkColorSpaceXform
Fallback: Raster pipeline

BUG=skia:6021

Change-Id: I3f0e36b08b051ab4ba3949f04474f20d0c20a963
Reviewed-on: https://skia-review.googlesource.com/7435
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Matt Sarett <msarett@google.com>

move sk_memset16,32,64 out-of-line

This helps pick it out on profiles and when debugging.

This will have a negligible impact on speed in most cases (small memsets are always fast, large memsets dwarf function call overhead), and clients like Chrome that use link-time optimization should see exactly zero impact.

This does however affect code size: it makes libskia 43K smaller on my laptop.

Change-Id: I55bdc75b7429a63024ab3b7d55dc1f886bde102a
Reviewed-on: https://skia-review.googlesource.com/6589
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>

Make GrConvolutionEffect only support Gaussian kernels and rename.

Change-Id: Ia874ad5bacc550b7ecec579719242e3354dca34b
Reviewed-on: https://skia-review.googlesource.com/7432
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>

Remove SkTDict and calls from SkView.

R=reed@google.com

Change-Id: Iebfddf752a2ac210a3f5c69de8decd06d0c35617
Reviewed-on: https://skia-review.googlesource.com/7436
Commit-Queue: Herb Derby <herb@google.com>
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>

Do not embed glyph image data in SkGlyphCache.

Instead of allocating memory for glyph images eagerly by embedding
the memory in the glyph cache, allocate memory dynamically on need.

TBR=bungeman@google.com

BUG=chromium:684366

Change-Id: If32bbc4d2608c976b93868feb519dcfa1212ce59
Reviewed-on: https://skia-review.googlesource.com/7433
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Herb Derby <herb@google.com>

conservativelyContainsRect is asserting

BUG=skia:6156

Change-Id: Iea571f001e9a4c4e70313c1bd49fb4e3c598869a
Reviewed-on: https://skia-review.googlesource.com/7423
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>

Add F16 and gamma correct support to webp encoder

BUG=skia:

Change-Id: Ib788466fa1e2bed26e7ffd8f03bee42930eae1d6
Reviewed-on: https://skia-review.googlesource.com/7425
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>

restore bool version of clip-bounds

BUG=skia:

Change-Id: I94e35566cf5bcd250515c71a566dd79030e2acb4
Reviewed-on: https://skia-review.googlesource.com/7430
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Mike Reed <reed@google.com>

Reenable index8 dsts

This is causing problems with image-cacherator-from-ctable gm.

In that gm, we use an index8->index8 readPixels() to take
a subset.

I think we need to keep this as long as we support index8 in the
generator/cacherator.  Though it's really strange because
we don't really bother to verify that color tables are the same
and we don't try to overwrite the dst color table...

BUG=skia:

Change-Id: Icf008caa571c4b81a52a23b8b7d9226789626ef5
Reviewed-on: https://skia-review.googlesource.com/7424
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>

Make GrYUVEffect take GrTextureProxies

This opens the door for swapping all the effects over to taking GrTextureProxies.

Change-Id: I3b03ba93a68f9945c9a8fee008fd170ed57616eb
Reviewed-on: https://skia-review.googlesource.com/7344
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>

Update SkAndroidSDKCanvas.

Hope that's right... nothing but the Android roll builds this code.

Change-Id: Ib86099deaaa5a67a480d6dc0eece8552e58aae1c
Reviewed-on: https://skia-review.googlesource.com/7428
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Temporarily restore colorcube bench and GM files.

Just the usual... we don't generate .bp for DM or nanobench yet.

See https://skia-review.googlesource.com/c/7363/ and
https://googleplex-android-review.git.corp.google.com/#/c/1819530/-1..1

Change-Id: If0098ed8a71a170d5efa773495d5bbf02a1e13fc
Reviewed-on: https://skia-review.googlesource.com/7427
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Fix comparison that overflows for addresses near uint max.
- Fix Assert

TBR=mtklein@google.com
BUG=chromium:683578

Change-Id: Iba503d1febace367c71f79a3b9accc0ec3e50f11
Reviewed-on: https://skia-review.googlesource.com/7418
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Herb Derby <herb@google.com>

make SkPath::conservativelyContainsRect consume degenerate segments

BUG=skia:

Change-Id: I3a39318bceaf6c95a50d84961d93af4ba62550e3
Reviewed-on: https://skia-review.googlesource.com/6900
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>

rename virtuals for clipbounds, and deprecate older bool/var-arg pattern

BUG=skia:

Change-Id: I08bcc2d0559e02838772538816b928e0716dd3aa
Reviewed-on: https://skia-review.googlesource.com/7412
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>

Revert "Update libwebp to 0.6.0-pre"

This reverts commit e2f6ffbf49671713356881af70a093e99e80bd31.

Reason for revert:

iOS build failures, see https://luci-milo.appspot.com/swarming/task/33e5e02cf8419d10/steps/build_iOSShell/0/stdout

Original change's description:
> Update libwebp to 0.6.0-pre
>
> Corresponds with Android change
> https://android-review.googlesource.com/#/c/326439/
>
> "this prerelease snapshot includes encoder and performance
> improvements"
>
> Update build file
> * Many files have been renamed from
>     src/<subdir>/<name>.c
>       to
>     src/<subdir>/<name>_<subdir>.c
>
> * Build new files (*_msa.c, *_neon.c, predictor_enc.c)
>
> This should fix issue 5876, which was caused by a compiler bug.
> With the added NEON implementation, we will no longer trigger
> the bug.
>
> BUG=skia:5876
>
> Change-Id: I0fcce4362ee70138547b1d23aa9ef537a4126e73
> Reviewed-on: https://skia-review.googlesource.com/7376
> Reviewed-by: Matt Sarett <msarett@google.com>
> Commit-Queue: Leon Scroggins <scroggo@google.com>
>

TBR=msarett@google.com,scroggo@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:5876

Change-Id: Ib81f63861d9f1c6936bd7f790b5e16a9544a4df0
Reviewed-on: https://skia-review.googlesource.com/7420
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>

Update libwebp to 0.6.0-pre

Corresponds with Android change
https://android-review.googlesource.com/#/c/326439/

"this prerelease snapshot includes encoder and performance
improvements"

Update build file
* Many files have been renamed from
    src/<subdir>/<name>.c
      to
    src/<subdir>/<name>_<subdir>.c

* Build new files (*_msa.c, *_neon.c, predictor_enc.c)

This should fix issue 5876, which was caused by a compiler bug.
With the added NEON implementation, we will no longer trigger
the bug.

BUG=skia:5876

Change-Id: I0fcce4362ee70138547b1d23aa9ef537a4126e73
Reviewed-on: https://skia-review.googlesource.com/7376
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>

Stop tracking whether GrFragmentProcessors read their input

Change-Id: Id5338901978b97289798eaef873527597bd8dfd6
Reviewed-on: https://skia-review.googlesource.com/7414
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>

Move shader register setup to SkRasterPipelineBlitter.

We've been seeding the initial values of our registers to x+0.5,y+0.5,
1,0, 0,0,0,0 (useful values for shaders to start with) in all pipelines.
This CL changes that to do so only when blitting, and only when we have
a shader.

The nicest part of this change is that SkRasterPipeline itself no longer
needs to have a concept of y, or what x means.  It just marches x
through [x,x+n), and the blitter handles y and layers the meaning of
"dst x coordinate" onto x.

This ought to make SkSplicer a little easier to work with too.

dm --src gm --config f16 srgb 565 all draws the same.

Change-Id: I69d8c1cc14a06e5dfdd6a7493364f43a18f8dec5
Reviewed-on: https://skia-review.googlesource.com/7353
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Revert "Fix comparison that overflows for addresses near uint max."

This reverts commit db8b8376b09cec9656b0ab864413b7911d81a84c.

Reason for revert:

Assertion failure, see https://luci-milo.appspot.com/swarming/task/33e5ae4d2bb25210/steps/dm/0/stdout

Original change's description:
> Fix comparison that overflows for addresses near uint max.
>
> BUG=chromium:683578
>
> Change-Id: I3f9b79eeeba3c68cccb72bd6423811c8ff8f2067
> Reviewed-on: https://skia-review.googlesource.com/7410
> Commit-Queue: Herb Derby <herb@google.com>
> Commit-Queue: Mike Klein <mtklein@chromium.org>
> Reviewed-by: Mike Klein <mtklein@chromium.org>
>

TBR=mtklein@chromium.org,mtklein@google.com,herb@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:683578

Change-Id: I198ea4c7209d060d0d15dfa3f6e555fa06e1a632
Reviewed-on: https://skia-review.googlesource.com/7415
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>

SkRasterPipeline: implement support for SkTableColorFilter

This adds and uses a byte_tables stage that converts to bytes, looks up
in the tables, then converts back to floats.

We treat this as color filter as pure math, not considering anything
colorspace related: no transfer functions, no gamut to change, etc.

Change-Id: If5fefc1bcef61a0fb0ae279002a0dd1547e429ea
Reviewed-on: https://skia-review.googlesource.com/7413
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

GN: add 32-bit and Release iOS builders.

And roll recipes to the point where they pass target_cpu to GN,
adding x86_64 as an alias for x64.

CQ_INCLUDE_TRYBOTS=skia.primary:Build-Mac-Clang-arm64-Debug-GN_iOS,Build-Mac-Clang-arm64-Release-GN_iOS,Build-Mac-Clang-arm-Debug-GN_iOS,Build-Mac-Clang-arm-Release-GN_iOS

Change-Id: I1933d5803ec7f59f78576c5a7b16489362905a97
Reviewed-on: https://skia-review.googlesource.com/7403
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Fix comparison that overflows for addresses near uint max.

BUG=chromium:683578

Change-Id: I3f9b79eeeba3c68cccb72bd6423811c8ff8f2067
Reviewed-on: https://skia-review.googlesource.com/7410
Commit-Queue: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Update SkTDict to use SkArenaAlloc.

Change-Id: I7ac33faa59bcad25b0580193af965a8525eb18e7
Reviewed-on: https://skia-review.googlesource.com/7360
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Track clear counts in vulkan render passes

Though not required by spec, the validation layers have started spitting
out warnings if the clearValueCount in BeginRenderPass is greater than
the number of attachments actually being cleared. This just adds tracking
of the count to appease the validation layers.

BUG=skia:

Change-Id: Iac6500df3ed5ad3f5df5f045c6e533bb021857aa
Reviewed-on: https://skia-review.googlesource.com/7401
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>

Make SkWEBPImageEncoder use transform_scanline_procs

This refactor will make it simple to add F16 and
GammaCorrect support to the webp encoder.

BUG=skia:

Change-Id: I379ba2ff0b44d865847877322e86d0c138e727c8
Reviewed-on: https://skia-review.googlesource.com/7356
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>

Reland "Refactor trimming logic for read/writePixels()"

Original CL: https://skia-review.googlesource.com/c/7326/

(1) Move trimming logic into Bitmap/Pixmap level for
    raster.  Everything goes through here, so we'll
    only do the work once.
(2) This means it also goes to GPU level.
(3) Always use SkReadPixelsRec rather than inlining
    the logic.
(4) Create an SkWritePixelsRec to encapsulate write
    trimming.
(5) Disabled kIndex8 as a dst - always.

CQ_INCLUDE_TRYBOTS=skia.primary:Perf-Ubuntu-Clang-GCE-CPU-AVX2-x86_64-Debug

BUG=skia:6021

Change-Id: I25a964e3c610c4e36d195a255e2150657baec649
Reviewed-on: https://skia-review.googlesource.com/7404
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>

Support fCapHeight in SkScalerContext_DW

BUG=chromium:681857

Change-Id: Ia97354d7798eb685351a98ff559bcf176a67af18
Reviewed-on: https://skia-review.googlesource.com/7380
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>

change clip-bounds getters to always return the rect
(actually fixes undefined result in getClipBounds)

future CLs
- update all callers to new apis
- move/rename virtuals

BUG=skia:

DOCS_PREVIEW= https://skia.org/?cl=7400

Change-Id: I45b93014e915c0d1c36d97d948c9ac8931f23258
Reviewed-on: https://skia-review.googlesource.com/7400
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Mike Reed <reed@google.com>

Remove public APIs that use SkDataTable.

Change-Id: Ibb8f987efc585713a6f8061c72fb416d07be13b1
Reviewed-on: https://skia-review.googlesource.com/7352
Reviewed-by: Mike Reed <reed@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>

Fix autoroller instructions

BUG=skia:6065

NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=7377

Change-Id: Iefd6af3104f4aea2f2ae93c556b6b429e54ebff4
Reviewed-on: https://skia-review.googlesource.com/7377
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>

Revert "Refactor trimming logic for read/writePixels()"

This reverts commit 977f64cbfad1ecd7fd4b1231c694c7e828fda1f0.

Reason for revert: Triggering nanobench asserts

Original change's description:
> Refactor trimming logic for read/writePixels()
>
> (1) Move trimming logic into Bitmap/Pixmap level for
>     raster.  Everything goes through here, so we'll
>     only do the work once.
> (2) This means it also goes to GPU level.
> (3) Always use SkReadPixelsRec rather than inlining
>     the logic.
> (4) Create an SkWritePixelsRec to encapsulate write
>     trimming.
> (5) Disabled kIndex8 as a dst - always.
>
> BUG=skia:6021
>
> Change-Id: I748f50c3b726f7c6de5462e2b1ccb54bc387a510
> Reviewed-on: https://skia-review.googlesource.com/7326
> Reviewed-by: Mike Reed <reed@google.com>
> Commit-Queue: Matt Sarett <msarett@google.com>
>

TBR=msarett@google.com,brianosman@google.com,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:6021

Change-Id: If9aacc6ce8b20e3dfe8a0f22ebca653f28356175
Reviewed-on: https://skia-review.googlesource.com/7379
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>

Allow conversion from non-opaque to opaque

BUG:683787

Change-Id: I1b78cc8d1b5d3917a2a952da036b93022e99e053
Reviewed-on: https://skia-review.googlesource.com/7374
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>

Revert "offset angle check edge in common"

This reverts commit d2eb581ebc8f8009e80cccccd74d5b341ef5bd5b.

Reason for revert: broke Google3 MSAN run of dm

Original change's description:
> offset angle check edge in common
>
> When curves cross, their intersection points may be nearby, but not exactly the same.
> Sort the angles formed by the crossing curves when all angles don't have the same
> origin.
>
> This sets up the framework to solve test case that currently fail (e.g., joel6) but
> does not fix all related test cases (e.g., joel9).
>
> All older existing test cases, including extended tests, pass.
>
> Rework the test framework to better report when tests expected to produce failing
> results now pass.
>
> Add new point and vector operations to support offset angles.
>
> TBR=reed@google.com
> BUG=skia:6041
>
> Change-Id: I67c651ded0a25e99ad93d55d6a35109b3ee3698e
> Reviewed-on: https://skia-review.googlesource.com/6624
> Commit-Queue: Cary Clark <caryclark@google.com>
> Reviewed-by: Cary Clark <caryclark@google.com>
>

TBR=caryclark@google.com,reviews@skia.org
# Not skipping CQ checks because original CL landed > 1 day ago.
BUG=skia:6041

Change-Id: I43db0808522ac44aceeb4f70e296167ea84a3663
Reviewed-on: https://skia-review.googlesource.com/7373
Commit-Queue: Cary Clark <caryclark@google.com>
Reviewed-by: Cary Clark <caryclark@google.com>

Get rid of level of indirection, noticed by VC++ 2017 bug

When building Chrome with VC++ 2017 the QuickFDot6Inverse::table pointer
is initialized to zero. This is certainly a bug, and it has been
reported to the VC++ team. But, the table pointer appears to be
unnecessary. Changing the code to reference the array directly should
give identical or perhaps even better code.

The change to the array indexing code is as follows:

- return table[x];
+ gFDot6INVERSE[kInverseTableSize + x];

This looks like a step backwards, but it isn't. 'table' is a pointer.
So, by default the compiler will load this pointer and then load the
value from the array. gFDot6INVERSE is an array, not a pointer. This
means that the adding of kInverseTableSize is trivially done at
compile time, and there is no loading of a pointer - only one data
segment memory access is needed.

The compiler may have realized that this optimization was ready before,
but now it is more trivial. And, this works around the VC++ 2017 bug so
that Chrome with VC++ 2017 will launch.

https://connect.microsoft.com/VisualStudio/feedback/details/3119337
BUG=683729

Change-Id: Iad443b59a70af83b39260e244e3242e782fafdbb
Reviewed-on: https://skia-review.googlesource.com/7284
Reviewed-by: Yuqian Li <liyuqian@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>

Refactor trimming logic for read/writePixels()

(1) Move trimming logic into Bitmap/Pixmap level for
    raster.  Everything goes through here, so we'll
    only do the work once.
(2) This means it also goes to GPU level.
(3) Always use SkReadPixelsRec rather than inlining
    the logic.
(4) Create an SkWritePixelsRec to encapsulate write
    trimming.
(5) Disabled kIndex8 as a dst - always.

BUG=skia:6021

Change-Id: I748f50c3b726f7c6de5462e2b1ccb54bc387a510
Reviewed-on: https://skia-review.googlesource.com/7326
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>

Fix performance issue and compute intersections more often.

(This CL also turned on Analytic AA for concave paths by removing SK_SUPPORT_LEGACY_AAA flag.)

Performance:
The SK_ALWAYS_INLINE was restored because it could bring 30%-50% speedup
in certain convex cases (e.g., fill_big_triangle). We also have to
reduce the number of branchings in the concave code path to enable such
speedup. (Although the speedup is for convex cases. The assembly code is
so strange...)

Intersection:
Previously, the criterion is too loose and that caused some bad pixels
(mostly unnoticeable by human eyes without magnifying). For example,
pixel (198, 222) of
https://gold.skia.org/detail?test=parsedpaths&digest=979e81de6f7b3f9e7e8dc810e31cad8d

BUG=skia:

Change-Id: I5e8191865c3df625f895cd4588c67c283fcbeaec
Reviewed-on: https://skia-review.googlesource.com/7318
Reviewed-by: Cary Clark <caryclark@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>

Look for GenerateMipMapEXT

Change-Id: I1ced904f0929653aa9bac09fee86607533e28be6
Reviewed-on: https://skia-review.googlesource.com/7359
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>

SkSplicer: favor 1.0f as a constant.

1.0f is the most commonly used constant, so we put it first in the
constant struct.  This usually cuts an instruction off loading it.

Change-Id: Ie54a9482aa3e90e9e5d8fcaa40dbd353af0fa1ca
Reviewed-on: https://skia-review.googlesource.com/7367
Reviewed-by: Mike Klein <mtklein@chromium.org>

Revert "Try /MTd on Debug Windows builds."

This reverts commit 58ce2bc51a318e53e3e6f1ef297c5adc775b0fec.

Reason for revert: warnings / errors when building ANGLE.

Original change's description:
> Try /MTd on Debug Windows builds.
>
> The people clamor for it!  I doubt we'll care in our testing.
>
> BUG=skia:5928
>
> Change-Id: I80d3948fd1f5d0b956c308bc3a6183b8660575ae
> Reviewed-on: https://skia-review.googlesource.com/7364
> Commit-Queue: Mike Klein <mtklein@chromium.org>
> Commit-Queue: Ben Wagner <bungeman@google.com>
> Reviewed-by: Ben Wagner <bungeman@google.com>
>

TBR=mtklein@chromium.org,bungeman@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:5928

Change-Id: Id6ab3de7ebb5394749233429bada2f5244ff2978
Reviewed-on: https://skia-review.googlesource.com/7372
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

SkSplicer: move armv7 ABI settings into build_stages.py.

This just tidies up the .cpp files a bit, and makes it easier to make
sure all exported functions use the aapcs-vfp callig convention, which
hard-float implies.

As a small simplification, fold -march=armv7-a into --target.

No generated code changes.

Change-Id: I2694970a6e48bd69c41dd280a44ddd0029e52ae8
Reviewed-on: https://skia-review.googlesource.com/7371
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Try /MTd on Debug Windows builds.

The people clamor for it!  I doubt we'll care in our testing.

BUG=skia:5928

Change-Id: I80d3948fd1f5d0b956c308bc3a6183b8660575ae
Reviewed-on: https://skia-review.googlesource.com/7364
Commit-Queue: Mike Klein <mtklein@chromium.org>
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>

mild tweaks to build_stages.py to run on my desktop

Change-Id: Ie193e5bf576b8d3ddbd4c99d3707e3211557cc01
Reviewed-on: https://skia-review.googlesource.com/7368
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Remove SkColorCubeFilter.  It is unused.

Change-Id: Iec5fc759e331de24caea1347f9510917260d379b
Reviewed-on: https://skia-review.googlesource.com/7363
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

minor cleanups in newly privatized device headers

BUG=skia:

Change-Id: Id6ebd54f404b5f0869d96368b98957ac10a2083f
NOTREECHECKS=True
Reviewed-on: https://skia-review.googlesource.com/7369
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>

clipstack is now private

BUG=skia:

Change-Id: I91e601d59eb5d46de17c280898ca01fcfab89e5c
NOTREECHECKS=True
Reviewed-on: https://skia-review.googlesource.com/7362
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>

Remove SkImage::MakeTextureFromPixmap

Another step towards reducing the number of texture upload paths.

BUG=skia:

Change-Id: Ica185e7334f52dc9ebf87e21fe6f60589ef87bb3
Reviewed-on: https://skia-review.googlesource.com/7346
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>

Replace raster pipeline nextafter() calls with SkNu nudging

(courtesy of mtklein@)

  nanobench -m gradient_linear_clamp\$ --config f16 --ms 2000 -q

Before: 753.66
After:  658.69

R=mtklein@google.com,herb@google.com

CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD

Change-Id: Ie1442da340f6cfc9aef65bec1f114c0e5db89fcb
Reviewed-on: https://skia-review.googlesource.com/7351
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Florin Malita <fmalita@chromium.org>

Revert "Revert "Make it possible to query GrXPFactory for dst texture without GrPipelineAnalysis.""

This reverts commit 3329cceab5feca230df1caee16be045249228dc1.

Reason for revert: Bot failures are unrelated to the original change.

Change-Id: I21b5927dc4384a25930bdefe16e57bcc9276ffa4
Reviewed-on: https://skia-review.googlesource.com/7347
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>

Force Index8 to N32 in deferred texture path.

Ganesh doesn't support Index8, because GPUs don't really
support it. So serializing it in the deferred blob just
meant doing an expansion before upload. This forces that
to happen when we build the blob. It also paves the way
for removing the last usage of MakeTextureFromPixmap,
which I'd like to remove.

BUG=skia:

Change-Id: I7c0292098d71e2f8ec1f9910e1234761822ce957
Reviewed-on: https://skia-review.googlesource.com/7340
Reviewed-by: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>

Revert "Make it possible to query GrXPFactory for dst texture without GrPipelineAnalysis."

This reverts commit f833215420847565b4c9945aebdc2e7ae182937f.

Reason for revert: <INSERT REASONING HERE>

Original change's description:
> Make it possible to query GrXPFactory for dst texture without GrPipelineAnalysis.
>
> Change-Id: I8c140eb4e3e5f2d21ecbf8f8f3c8533dc7f50e7c
> Reviewed-on: https://skia-review.googlesource.com/7316
> Commit-Queue: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Greg Daniel <egdaniel@google.com>
>

TBR=egdaniel@google.com,bsalomon@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: I790afb9a01422cb4c2d3a4be4ecd20e8c4466b29
Reviewed-on: https://skia-review.googlesource.com/7342
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>