More win64 warning fixes.

TBR=bsalomon@google.com
NOTREECHECKS=true

Review URL: https://codereview.chromium.org/887713002

Fix win64 warnings.

TBR=joshualitt@google.com

NOTREECHECKS=true

Review URL: https://codereview.chromium.org/884253004

skia: blend32_16_row for neon version

This includes blend32_16_row neon implementation
for aarch32 and aarch64.

For performance,
blend32_16_row is called in following tests in nanobench.
 - Xfermode_SrcOver
 - tablebench
 - rotated_rects_bw_alternating_transparent_and_opaque_srcover
 - rotated_rects_bw_changing_transparent_srcover
 - rotated_rects_bw_same_transparent_srcover
 - luma_colorfilter_large
 - luma_colorfilter_small
 - chart_bw

I can see perf increase in following two tests, especially. For others, looks
similar.
For each, I tried to run two times.

1) Xfermode_SrcOver
<org>
 - D/skia    ( 2000):    3M        57      17.3µs  17.4µs  17.4µs  17.7µs  1%
  █▃▂▃▂▂▂▁▃▂      565     Xfermode_SrcOver
 - D/skia    ( 1915):    3M        70      13.5µs  16.9µs  16.7µs  18.8µs  9%
  ▆█▄▅█▁▅▅▆▄      565     Xfermode_SrcOver

<new>
 - D/skia    ( 2000):    3M        8       11.6µs  11.8µs  12.1µs  14.4µs  7%
  ▃█▁▁▂▁▁▁▂▂      565     Xfermode_SrcOver
 - D/skia    ( 2004):    3M        62      10.3µs  12.9µs  13µs    15.2µs  11%
  █▅▅▆▁▅▅▅▇▃      565     Xfermode_SrcOver

2)
luma_colorfilter_large
<org>
 - D/skia    ( 2000):  159M        8       136µs   136µs   136µs   139µs   1%
  █▃▁▂▁▁▁▁▁▁      565     luma_colorfilter_large
 - D/skia    ( 1915):  158M        2       135µs   177µs   182µs   269µs   22%
  ▆▃█▁▁▃▃▃▃▃      565     luma_colorfilter_large

<new>
 - D/skia    ( 2000):  157M        5       84.2µs  85.3µs  87.5µs  110µs   9%
  █▁▂▁▁▁▁▁▁▁      565     luma_colorfilter_large
 - D/skia    ( 2004):  159M        6       84.7µs  110µs   112µs   144µs   18%
  █▄▇▁▁▄▃▄▄▆      565     luma_colorfilter_large

Review URL: https://codereview.chromium.org/847363002

Revert of Add device space "nudge" to gpu draws (patchset #5 id:70001 of https://codereview.chromium.org/877473005/)

Reason for revert:
Chrome pixel test :(

Original issue's description:
> Add device space "nudge" to gpu draws
>
> This CL nudges all the GPU draws and clips slightly to match raster's round behavior for BW draws. We assume the effect will be negligible and do it for AA draws too.
>
> BUG=423834
>
> Committed: https://skia.googlesource.com/skia/+/2d55d07501c56310f97d2092d789a2bc9fa01b78

TBR=bsalomon@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=423834

Review URL: https://codereview.chromium.org/890433003

Revert of Remove 'f' from 0.05f in shader code (patchset #1 id:1 of https://codereview.chromium.org/888483002/)

Reason for revert:
Chrome pixel test

Original issue's description:
> Remove 'f' from 0.05f in shader code
>
> TBR=bsalomon@google.com
> NOTREECHECKS=true
> NOTRY=true
>
> Committed: https://skia.googlesource.com/skia/+/1726997861fac8daa8213d1a51d5c8fbe44428d1

TBR=bsalomon@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/890433002

Revert of DM::SKPSrc::size() reports correct size. (patchset #3 id:40001 of https://codereview.chromium.org/863243005/)

Reason for revert:
Still no good on Chrome OS bot:

http://build.chromium.org/p/client.skia/builders/Test-ChromeOS-Alex-GMA3150-x86-Release/builds/628/steps/dm/logs/stdio

Original issue's description:
> DM::SKPSrc::size() reports correct size.
>
> Also, DM::GPUSink and DM::RasterSink crop DM::Src::size() to 2048x2048.
>
> Motivation:
>   Improve PDF testing by printing the entire SKP.
>
> Source: http://crrev.com/863243004
>
> BUG=skia:3365
>
> Committed: https://skia.googlesource.com/skia/+/441b10eac09a1f44983e35da827a6b438a409e63
>
> CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu12-ShuttleA-GTX660-x86-Release-Trybot,Test-ChromeOS-Alex-GMA3150-x86-Release-Trybot
>
> Committed: https://skia.googlesource.com/skia/+/d4dd58e43ca4551531ad6a9f54bfc5632ea45a80

TBR=halcanary@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:3365

Review URL: https://codereview.chromium.org/886543005

Remove 'f' from 0.05f in shader code

TBR=bsalomon@google.com
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/888483002

Add device space "nudge" to gpu draws

This CL nudges all the GPU draws and clips slightly to match raster's round behavior for BW draws. We assume the effect will be negligible and do it for AA draws too.

BUG=423834

Review URL: https://codereview.chromium.org/877473005

authors update- add herb and reorganize

BUG=skia:

TBR=reed@google.com

NOTRY=true

Review URL: https://codereview.chromium.org/881993002

Make char hash dynamic when needed.

BUG=skia:

Review URL: https://codereview.chromium.org/880383002

DM::SKPSrc::size() reports correct size.

Also, DM::GPUSink and DM::RasterSink crop DM::Src::size() to 2048x2048.

Motivation:
  Improve PDF testing by printing the entire SKP.

Source: http://crrev.com/863243004

BUG=skia:3365

Committed: https://skia.googlesource.com/skia/+/441b10eac09a1f44983e35da827a6b438a409e63

CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu12-ShuttleA-GTX660-x86-Release-Trybot

Review URL: https://codereview.chromium.org/863243005

Use murmur3 finisher to improve font hash efficiency.
Add dump() method to inspect glyphcache strikes.

Murmur addition improves hash efficient roughly 50%

BUG=skia:

Review URL: https://codereview.chromium.org/877113002

Fixed clusterfuzz issue

BUG=448423

Review URL: https://codereview.chromium.org/881423002

Use distance fields for glyphs > 256 pt, before switching to paths.

BUG=452313

Review URL: https://codereview.chromium.org/862403004

GrBatchPrototype

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/d15e4e45374275c045572b304c229237c4a82be4

Committed: https://skia.googlesource.com/skia/+/d5a7db4a867c7e6ccf8451a053d987b470099198

Review URL: https://codereview.chromium.org/845103005

Revert of DM::SKPSrc::size() reports correct size. (patchset #1 id:1 of https://codereview.chromium.org/863243005/)

Reason for revert:
OOM on 32-bit machines.

Original issue's description:
> DM::SKPSrc::size() reports correct size.
>
> Also, DM::GPUSink and DM::RasterSink crop DM::Src::size() to 2048x2048.
>
> Motivation:
>   Improve PDF testing by printing the entire SKP.
>
> Source: http://crrev.com/863243004
>
> BUG=skia:3365
>
> Committed: https://skia.googlesource.com/skia/+/441b10eac09a1f44983e35da827a6b438a409e63

TBR=halcanary@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:3365

Review URL: https://codereview.chromium.org/884743003

DM::PDFSink::draw excercises multi-page pdf

BUG=skia:3365

Review URL: https://codereview.chromium.org/881343002

add more checks for computing clamp counts, remove dead code

BUG=448299

Review URL: https://codereview.chromium.org/886473003

Fix wrapped content keys for npot textures.

Review URL: https://codereview.chromium.org/879193002

Fold gmtoskp into DM, as --src gm --config skp.

BUG=skia:

Review URL: https://codereview.chromium.org/885733002

DM::SKPSrc::size() reports correct size.

Also, DM::GPUSink and DM::RasterSink crop DM::Src::size() to 2048x2048.

Motivation:
  Improve PDF testing by printing the entire SKP.

Source: http://crrev.com/863243004

BUG=skia:3365

Review URL: https://codereview.chromium.org/863243005

dstread gm

TBR=
BUG=skia:

Review URL: https://codereview.chromium.org/883053002

Add a flag to flush the canvases during SkMultiPictureDraw::draw().

This is necessary for multisampling, so that each multisampled render
target resolves before Chrome's compositor attempts to draw the
texture.

BUG=skia:

Review URL: https://codereview.chromium.org/878653004

Conservative SkTextBlob bounds.

Compute cheaper/more conservative text blob bounds based on the typeface
maximum glyph bbox.

BUG=chromium:451401
R=reed@google.com,bungeman@google.com

Review URL: https://codereview.chromium.org/886473002

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=

Review URL: https://codereview.chromium.org/879203002

Add a script to fetch the latest SKPs.

I keep forgetting how best to do this.

NOTRY=true
NOTREECHECKS=true

Review URL: https://codereview.chromium.org/880283002

add a paranoid assert

NOTREECHECKS=true

BUG=chromium:399842

Review URL: https://codereview.chromium.org/881253003

Revert of GrBatchPrototype (patchset #32 id:630001 of https://codereview.chromium.org/845103005/)

Reason for revert:
One last try to fix mac perf regression

Original issue's description:
> GrBatchPrototype
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/d15e4e45374275c045572b304c229237c4a82be4
>
> Committed: https://skia.googlesource.com/skia/+/d5a7db4a867c7e6ccf8451a053d987b470099198

TBR=bsalomon@google.com,kkinnunen@nvidia.com,joshualitt@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/877393002

whitespace

NOTREECHECKS=True
NOTRY=True
TBR=
BUG=skia:

Review URL: https://codereview.chromium.org/885443003

Revert of patch from issue 885453002 at patchset 20001 (crrev.com/885453002#ps20001) (patchset #1 id:1 of https://codereview.chromium.org/881953002/)

Reason for revert:
==32435==ERROR: AddressSanitizer: alloc-dealloc-mismatch (operator new [] vs operator delete) on 0x621000d8cd00

Lots of info here:
http://build.chromium.org/p/client.skia/builders/Test-Ubuntu13.10-GCE-NoGPU-x86_64-Debug-ASAN/builds/1198/steps/dm/logs/stdio

Original issue's description:
> patch from issue 885453002 at patchset 20001 (http://crrev.com/885453002#ps20001)
>
> Make the char cache dynamic in SkGlyphCache
> because it is rarely used.
>
> Landing on behalf of Herb.
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/95faa61d63a6f62916f6f7be58c4624da8357e3b

TBR=mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/881023003

GrBatchPrototype

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/d15e4e45374275c045572b304c229237c4a82be4

Review URL: https://codereview.chromium.org/845103005

patch from issue 885453002 at patchset 20001 (crrev.com/885453002#ps20001)

Make the char cache dynamic in SkGlyphCache
because it is rarely used.

Landing on behalf of Herb.

BUG=skia:

Review URL: https://codereview.chromium.org/881953002

add -r to DM

$ out/Debug/dm -w good
$ out/Debug/dm -r good -w bad && echo "hooray no diffs!"

BUG=skia:

Review URL: https://codereview.chromium.org/863093003

SSE4 opaque blend using intrinsics instead of assembly.

Since we had such a hard time with the assembly versions of this blit (to the
point that we have them completely disabled everywhere), I thought I'd take
a shot at writing a version of the blit using intrinsics.

The key feature of SSE4 we're exploiting is that we can use ptest (_mm_test*)
to skip the blend when the 16 src pixels we consider each loop are all opaque
or all transparent.  _mm_shuffle_epi8 from SSSE3 also lends a hand to extract
all those alphas.

It's worth looking to see if we can backport this type of logic to SSE2 using
_mm_movemask_epi8, or up to 32 pixels at a time using AVX.

My local performance testing doesn't show this to be an unambiguous win
(there are probably microbenchmarks and SKPs where we'd be better off just
powering through the blend rather than looking at alphas), but the potential
does seem tantalizing enough to let skiaperf vet it on the bots.  (< 1.0x is a win.)

DM says it draws pixel perfect compare to the old code.

Microbenchmarks:
               bitmap_RGBA_8888_A_source_stripes_two   14us -> 14.4us 1.03x
             bitmap_RGBA_8888_A_source_stripes_three 14.3us -> 14.5us 1.01x
                       bitmap_RGBA_8888_scale_bilerp 61.9us -> 62.2us 1.01x
bitmap_RGBA_8888_update_volatile_scale_rotate_bilerp  102us ->  101us 0.99x
                bitmap_RGBA_8888_scale_rotate_bilerp  103us ->  101us 0.99x
                              bitmap_RGBA_8888_scale 18.4us -> 18.2us 0.99x
             bitmap_RGBA_8888_A_scale_rotate_bicubic   71us ->   70us 0.99x
         bitmap_RGBA_8888_update_scale_rotate_bilerp  103us ->  101us 0.99x
              bitmap_RGBA_8888_A_scale_rotate_bilerp  112us ->  109us 0.98x
                    bitmap_RGBA_8888_update_volatile 5.72us -> 5.58us 0.98x
                                    bitmap_RGBA_8888 5.73us -> 5.58us 0.97x
                             bitmap_RGBA_8888_update 5.78us ->  5.6us 0.97x
                     bitmap_RGBA_8888_A_scale_bilerp 70.7us ->   68us 0.96x
                    bitmap_RGBA_8888_A_scale_bicubic 23.7us -> 21.8us 0.92x
                                  bitmap_RGBA_8888_A 13.9us -> 10.9us 0.78x
                    bitmap_RGBA_8888_A_source_opaque   14us -> 6.29us 0.45x
               bitmap_RGBA_8888_A_source_transparent   14us -> 3.65us 0.26x

Running over our ~70 SKP web page captures, this looks like we spend 0.7x
the time in S32A_Opaque_BlitRow compared to the SSE2 version, which should
be a decent predictor of real-world impact.

BUG=chromium:399842

Committed: https://skia.googlesource.com/skia/+/04bc91b972417038fecfa87c484771eac2b9b785

CQ_EXTRA_TRYBOTS=client.skia:Test-Mac10.6-MacMini4.1-GeForce320M-x86_64-Release-Trybot

Review URL: https://codereview.chromium.org/874863002

Remove unused methods from SkScalerContext.

The methods getLocalMatrixWithoutTextSize and
getSingleMatrixWithoutTextSize on SkScalerContext were added as a
temporary measure for CoreText issues. Now that the CoreText
SkScalerContext is using other means to fix these issues more completely,
remove these now unused methods.

Review URL: https://codereview.chromium.org/883833002

SkFontHost_FreeType takes advantage of SkStreamAsset.

With recent changes, SkTypeface now deals in SkStreamAsset instead of
SkStream. Take advantage of this for performance with FreeType.

Review URL: https://codereview.chromium.org/882763002

Fix GPU resource cache related assertions.

Review URL: https://codereview.chromium.org/879963003

Setup Android framework builds to use the appropriate shared lib defines.

Review URL: https://codereview.chromium.org/864043005

sk_tool_utils::draw_checkerboard uses SkXfermode::kSrc_Mode to fix valgrind error.

Review URL: https://codereview.chromium.org/877103002

Use highp for distance field texture coordinates.

Addresses issues with wavy text (probably due to low-precision aliasing)

BUG=429080

Review URL: https://codereview.chromium.org/879603003

fix gm to not rely on SkColor's swizzle == SkPMColor's

BUG=skia:3361

Review URL: https://codereview.chromium.org/873983009

Revert of GrBatchPrototype (patchset #30 id:570001 of https://codereview.chromium.org/845103005/)

Reason for revert:
creates large performance regression

Original issue's description:
> GrBatchPrototype
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/d15e4e45374275c045572b304c229237c4a82be4

TBR=bsalomon@google.com,joshualitt@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/862823004

Add ClipDrawMatch SampleApp slide

This slide can be used to find and diagnose discrepancies between BW clipping and drawing.

BUG=skia:423834

Review URL: https://codereview.chromium.org/872363003

SkTypeface to use SkStreamAsset.

SkTypeface already requires typeface streams to support SkStreamAsset
in practice, and in practice all users are already supplying them.

Review URL: https://codereview.chromium.org/869763002

speedup mipmap building

mipmap_build benchmark:

before: 3.36ms
after:    2.20ms

BUG=skia:

Review URL: https://codereview.chromium.org/873393002

Make stencil buffers uncached for uncached render target textures

Make new stencil buffers of uncached render target textures not affect the
cache budgets. This is consistent with render buffer storage of uncached
render target textures.

Affects only newly created stencil buffers. An uncached render target
might still receive a cached stencil buffer if such is available from
cache.

BUG=skia:3119
BUG=skia:3301

Review URL: https://codereview.chromium.org/859013002

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=

Review URL: https://codereview.chromium.org/882613002

Split src/opts source lists out of opts.gyp.

This should make it easier to keep our opts.gyp in sync with Chrome's GYP and GN.

BUG=skia:

Landing this without review as a mega-tryjob.
TBR=reed@google.com

Committed: https://skia.googlesource.com/skia/+/c98fe3aa4f8c97c462c0eb6d9106fc37e48d7f82

Review URL: https://codereview.chromium.org/870353003

Revert of Split src/opts source lists out of opts.gyp. (patchset #1 id:1 of https://codereview.chromium.org/870353003/)

Reason for revert:
Android Makefiles broken

Original issue's description:
> Split src/opts source lists out of opts.gyp.
>
> This should make it easier to keep our opts.gyp in sync with Chrome's GYP and GN.
>
> BUG=skia:
>
> Landing this without review as a mega-tryjob.
> TBR=reed@google.com
>
> Committed: https://skia.googlesource.com/skia/+/c98fe3aa4f8c97c462c0eb6d9106fc37e48d7f82

TBR=mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/880783002

Split src/opts source lists out of opts.gyp.

This should make it easier to keep our opts.gyp in sync with Chrome's GYP and GN.

BUG=skia:

Landing this without review as a mega-tryjob.
TBR=reed@google.com

Review URL: https://codereview.chromium.org/870353003

Revert of SSE4 opaque blend using intrinsics instead of assembly. (patchset #14 id:260001 of https://codereview.chromium.org/874863002/)

Reason for revert:
This kills Mac 10.6 bots.

FAILED: c++ -MMD -MF obj/src/opts/opts_sse4.SkBlitRow_opts_SSE4.o.d -DSK_INTERNAL -DSK_GAMMA_SRGB -DSK_GAMMA_APPLY_TO_A8 -DSK_SCALAR_TO_FLOAT_EXCLUDED -DSK_ALLOW_STATIC_GLOBAL_INITIALIZERS=1 -DSK_SUPPORT_GPU=1 -DSK_SUPPORT_OPENCL=0 -DSK_FORCE_DISTANCE_FIELD_TEXT=0 -DSK_BUILD_FOR_MAC -DSK_CRASH_HANDLER -DSK_DEVELOPER=1 -I../../src/core -I../../src/utils -I../../include/c -I../../include/config -I../../include/core -I../../include/pathops -I../../include/pipe -I../../include/utils/mac -I../../include/effects -O0 -gdwarf-2 -mmacosx-version-min=10.6 -arch x86_64 -mssse3 -Wall -Wextra -Winit-self -Wpointer-arith -Wsign-compare -Wno-unused-parameter -Wno-invalid-offsetof -msse4.1  -c ../../src/opts/SkBlitRow_opts_SSE4.cpp -o obj/src/opts/opts_sse4.SkBlitRow_opts_SSE4.o
../../src/opts/SkBlitRow_opts_SSE4.cpp:15:27: warning: x86intrin.h: No such file or directory
../../src/opts/SkBlitRow_opts_SSE4.cpp: In function 'void S32A_Opaque_BlitRow32_SSE4(SkPMColor*, const SkPMColor*, int, U8CPU)':
../../src/opts/SkBlitRow_opts_SSE4.cpp:40: error: '_mm_testz_si128' was not declared in this scope
../../src/opts/SkBlitRow_opts_SSE4.cpp:45: error: '_mm_testc_si128' was not declared in this scope

Original issue's description:
> SSE4 opaque blend using intrinsics instead of assembly.
>
> Since we had such a hard time with the assembly versions of this blit (to the
> point that we have them completely disabled everywhere), I thought I'd take
> a shot at writing a version of the blit using intrinsics.
>
> The key feature of SSE4 we're exploiting is that we can use ptest (_mm_test*)
> to skip the blend when the 16 src pixels we consider each loop are all opaque
> or all transparent.  _mm_shuffle_epi8 from SSSE3 also lends a hand to extract
> all those alphas.
>
> It's worth looking to see if we can backport this type of logic to SSE2 using
> _mm_movemask_epi8, or up to 32 pixels at a time using AVX.
>
> My local performance testing doesn't show this to be an unambiguous win
> (there are probably microbenchmarks and SKPs where we'd be better off just
> powering through the blend rather than looking at alphas), but the potential
> does seem tantalizing enough to let skiaperf vet it on the bots.  (< 1.0x is a win.)
>
> DM says it draws pixel perfect compare to the old code.
>
> Microbenchmarks:
>                bitmap_RGBA_8888_A_source_stripes_two   14us -> 14.4us 1.03x
>              bitmap_RGBA_8888_A_source_stripes_three 14.3us -> 14.5us 1.01x
>                        bitmap_RGBA_8888_scale_bilerp 61.9us -> 62.2us 1.01x
> bitmap_RGBA_8888_update_volatile_scale_rotate_bilerp  102us ->  101us 0.99x
>                 bitmap_RGBA_8888_scale_rotate_bilerp  103us ->  101us 0.99x
>                               bitmap_RGBA_8888_scale 18.4us -> 18.2us 0.99x
>              bitmap_RGBA_8888_A_scale_rotate_bicubic   71us ->   70us 0.99x
>          bitmap_RGBA_8888_update_scale_rotate_bilerp  103us ->  101us 0.99x
>               bitmap_RGBA_8888_A_scale_rotate_bilerp  112us ->  109us 0.98x
>                     bitmap_RGBA_8888_update_volatile 5.72us -> 5.58us 0.98x
>                                     bitmap_RGBA_8888 5.73us -> 5.58us 0.97x
>                              bitmap_RGBA_8888_update 5.78us ->  5.6us 0.97x
>                      bitmap_RGBA_8888_A_scale_bilerp 70.7us ->   68us 0.96x
>                     bitmap_RGBA_8888_A_scale_bicubic 23.7us -> 21.8us 0.92x
>                                   bitmap_RGBA_8888_A 13.9us -> 10.9us 0.78x
>                     bitmap_RGBA_8888_A_source_opaque   14us -> 6.29us 0.45x
>                bitmap_RGBA_8888_A_source_transparent   14us -> 3.65us 0.26x
>
> Running over our ~70 SKP web page captures, this looks like we spend 0.7x
> the time in S32A_Opaque_BlitRow compared to the SSE2 version, which should
> be a decent predictor of real-world impact.
>
> BUG=chromium:399842
>
> Committed: https://skia.googlesource.com/skia/+/04bc91b972417038fecfa87c484771eac2b9b785

TBR=henrik.smiding@intel.com,mtklein@google.com,herb@google.com,reed@google.com,thakis@chromium.org,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:399842

Review URL: https://codereview.chromium.org/874033004

Add sbix font to coloremoji gm.

Review URL: https://codereview.chromium.org/797043002

SSE4 opaque blend using intrinsics instead of assembly.

Since we had such a hard time with the assembly versions of this blit (to the
point that we have them completely disabled everywhere), I thought I'd take
a shot at writing a version of the blit using intrinsics.

The key feature of SSE4 we're exploiting is that we can use ptest (_mm_test*)
to skip the blend when the 16 src pixels we consider each loop are all opaque
or all transparent.  _mm_shuffle_epi8 from SSSE3 also lends a hand to extract
all those alphas.

It's worth looking to see if we can backport this type of logic to SSE2 using
_mm_movemask_epi8, or up to 32 pixels at a time using AVX.

My local performance testing doesn't show this to be an unambiguous win
(there are probably microbenchmarks and SKPs where we'd be better off just
powering through the blend rather than looking at alphas), but the potential
does seem tantalizing enough to let skiaperf vet it on the bots.  (< 1.0x is a win.)

DM says it draws pixel perfect compare to the old code.

Microbenchmarks:
               bitmap_RGBA_8888_A_source_stripes_two   14us -> 14.4us 1.03x
             bitmap_RGBA_8888_A_source_stripes_three 14.3us -> 14.5us 1.01x
                       bitmap_RGBA_8888_scale_bilerp 61.9us -> 62.2us 1.01x
bitmap_RGBA_8888_update_volatile_scale_rotate_bilerp  102us ->  101us 0.99x
                bitmap_RGBA_8888_scale_rotate_bilerp  103us ->  101us 0.99x
                              bitmap_RGBA_8888_scale 18.4us -> 18.2us 0.99x
             bitmap_RGBA_8888_A_scale_rotate_bicubic   71us ->   70us 0.99x
         bitmap_RGBA_8888_update_scale_rotate_bilerp  103us ->  101us 0.99x
              bitmap_RGBA_8888_A_scale_rotate_bilerp  112us ->  109us 0.98x
                    bitmap_RGBA_8888_update_volatile 5.72us -> 5.58us 0.98x
                                    bitmap_RGBA_8888 5.73us -> 5.58us 0.97x
                             bitmap_RGBA_8888_update 5.78us ->  5.6us 0.97x
                     bitmap_RGBA_8888_A_scale_bilerp 70.7us ->   68us 0.96x
                    bitmap_RGBA_8888_A_scale_bicubic 23.7us -> 21.8us 0.92x
                                  bitmap_RGBA_8888_A 13.9us -> 10.9us 0.78x
                    bitmap_RGBA_8888_A_source_opaque   14us -> 6.29us 0.45x
               bitmap_RGBA_8888_A_source_transparent   14us -> 3.65us 0.26x

Running over our ~70 SKP web page captures, this looks like we spend 0.7x
the time in S32A_Opaque_BlitRow compared to the SSE2 version, which should
be a decent predictor of real-world impact.

BUG=chromium:399842

Review URL: https://codereview.chromium.org/874863002

site/dev/contrib/directory <= https://status.skia.org/

NOTRY=true

Review URL: https://codereview.chromium.org/876873002

GrBatchPrototype

BUG=skia:

Review URL: https://codereview.chromium.org/845103005

Remove clear function from SkGPipe

Clear no longer exists.  (It's now a drawPaint call)

Review URL: https://codereview.chromium.org/881563002

s/sk_tools::DrawCheckerboard/sk_tool_utils::draw_checkerboard/

BUG=skia:

Review URL: https://codereview.chromium.org/873333004

site/dev/contrib/revert formatting and details

NOTRY=true

Review URL: https://codereview.chromium.org/873483004

add bench for building mipmaps

BUG=skia:
TBR=

Review URL: https://codereview.chromium.org/873293003

Fix Morphology effects sourcing outside of the crop rect.

BUG=skia:1766

Committed: https://skia.googlesource.com/skia/+/f6be925b5615f07039ce95c3433039694a8d1679

Review URL: https://codereview.chromium.org/781153002

discard scratch render targets when they are recycled.

Review URL: https://codereview.chromium.org/871373007

Alter gpu veto

This CL unifies the treatment of the dashed and concave paths.

Before:
TP 28 FP 15 TN 8 FN 3 IND 3

After:
TP 28 FP 18 TN 7 FN 2 IND 2

One of the TrueNegatives that became a FalsePositive was the motivation use case (the Chromium busy spinner).

Committed: https://skia.googlesource.com/skia/+/87a6a8e18c7d5bbc94f478b44c53dc0e0549f927

Review URL: https://codereview.chromium.org/875913002

check effective cache-size for fixed-budget caches

BUG=skia:

Review URL: https://codereview.chromium.org/876743002

Factor out checkerboard function in gm and sampleapp into tools.

Review URL: https://codereview.chromium.org/834303005

prettify site/user/sample/hello

NOTRY=true
TBR=jcgregorio@google.com

Review URL: https://codereview.chromium.org/874303002

Stop incorrectly setting the clip bit in GrCMM

Review URL: https://codereview.chromium.org/866403005

Collapse consecutive SkTableColorFilters

BUG=skia:1366

For the added bench, the collapsing makes the bench take:
- 70% of the time for CPU rendering of 3 consecutive matrix filters
- almost no change in the GPU rendering of the matrix filters
- 50% of the time for CPU and GPU rendering of 3 consecutive table filters

Review URL: https://codereview.chromium.org/776673002

prettify site/dev/design/pdftheory

NOTRY=true
TBR=jcgregorio@google.com

Review URL: https://codereview.chromium.org/868473006

Add docs on how to get a preview of markdown pages.

BUG=skia:

Review URL: https://codereview.chromium.org/878593002

https://fiddle.skia.org/

NOTRY=true

Review URL: https://codereview.chromium.org/868473005

Don't do a pointless << 0.

It's very common (universal?) that alpha is the top byte.
You'd hope the compiler would remove the left shift then,
but I've seen Clang just do a dumb left shift of zero. :(

BUG=skia:

Review URL: https://codereview.chromium.org/872243003

Revert of Alter gpu veto (patchset #1 id:1 of https://codereview.chromium.org/875913002/)

Reason for revert:
Failing tests

Original issue's description:
> Alter gpu veto
>
> This CL unifies the treatment of the dashed and concave paths.
>
> Before:
> TP 28 FP 15 TN 8 FN 3 IND 3
>
> After:
> TP 28 FP 18 TN 7 FN 2 IND 2
>
> One of the TrueNegatives that became a FalsePositive was the motivation use case (the Chromium busy spinner).
>
> Committed: https://skia.googlesource.com/skia/+/87a6a8e18c7d5bbc94f478b44c53dc0e0549f927

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

Review URL: https://codereview.chromium.org/879483003

Fix Chrome build

Switch SkShader's toString method to not be pure virtual due to derived classes in Chromium

TBR=bsalomon@google.com
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/875043002

Alter gpu veto

This CL unifies the treatment of the dashed and concave paths.

Before:
TP 28 FP 15 TN 8 FN 3 IND 3

After:
TP 28 FP 18 TN 7 FN 2 IND 2

One of the TrueNegatives that became a FalsePositive was the motivation use case (the Chromium busy spinner).

Review URL: https://codereview.chromium.org/875913002

Update references to skiaperf.com.

The new server is being run in perf.skia.org.

BUG=None
R=jcgregorio@google.com

Review URL: https://codereview.chromium.org/866943003

Add patheffects to debugger printout

TBR=bsalomon@google.com

Review URL: https://codereview.chromium.org/872043002

Fix Windows build

TBR=jvanverth@google.com
NOTRY=true
NOTREECHECKS=true

Review URL: https://codereview.chromium.org/876643002

Fold alpha to the inner savelayer in savelayer-savelayer-restore patterns

Fold alpha to the inner savelayer in savelayer-savelayer-restore
patterns such as this:

  SaveLayer (non-opaque)
    Save
      ClipRect
      SaveLayer
      Restore
    Restore
  Restore

Current blink generates these for example for SVG content such as this:

<path style="opacity:0.5 filter:url(#blur_filter)"/>

The outer save layer is due to the opacity and the inner one is due to
blur filter being implemented with picture image filter.

Reduces layers in desk_carsvg.skp testcase from 115 to 78.

BUG=skia:3119

Review URL: https://codereview.chromium.org/835973005

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=

Review URL: https://codereview.chromium.org/873163002

initial preroll api

BUG=skia:

Review URL: https://codereview.chromium.org/855473002

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=

Review URL: https://codereview.chromium.org/870353002

experimental/skp_to_pdf_md5 optionally also outputs pdf files

TBR=mtklein@google.com

Review URL: https://codereview.chromium.org/868333002

Don't leak image in Surface test.

==7023== 5,056 (896 direct, 4,160 indirect) bytes in 8 blocks are definitely lost in loss record 947 of 2,656
==7023==    at 0x4C2B1C7: operator new(unsigned long) (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==7023==    by 0x82CB99: SkNewImageFromBitmapTexture(SkBitmap const&, int, SkSurface::Budgeted) (SkImage_Gpu.cpp:64)
==7023==    by 0x81A1CF: SkSurface_Gpu::onNewImageSnapshot(SkSurface::Budgeted) (SkSurface_Gpu.cpp:50)
==7023==    by 0x694F9D: SkSurface::newImageSnapshot(SkSurface::Budgeted) (SkSurface_Base.h:92)
==7023==    by 0x55EA7C: test_Surface(skiatest::Reporter*, GrContextFactory*) (SurfaceTest.cpp:485)
==7023==    by 0x407CB2: run_test(skiatest::Test*) (DM.cpp:399)
==7023==    by 0x408382: run_enclave_and_gpu_tests(SkTArray<Task, false>*) (DM.cpp:411)
==7023==    by 0x68B71E: SkTaskGroup::wait() (SkTaskGroup.cpp:67)
==7023==    by 0x40934E: dm_main() (DM.cpp:455)
==7023==    by 0x409483: main (DM.cpp:477)

http://build.chromium.org/p/client.skia/builders/Test-Ubuntu12-ShuttleA-GTX550Ti-x86_64-Release-Valgrind/builds/276/steps/dm/logs/stdio

TBR=bsalomon@google.com

BUG=skia:

Review URL: https://codereview.chromium.org/872183002

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=

Review URL: https://codereview.chromium.org/871143002

a typo in SkResourceCacheTest

BUG=skia:

Review URL: https://codereview.chromium.org/868613002

Revert of Fix Morphology effects sourcing outside of the crop rect. (patchset #6 id:100001 of https://codereview.chromium.org/781153002/)

Reason for revert:
Looks like this is causing memory leaks:
http://build.chromium.org/p/client.skia/builders/Test-Ubuntu13.10-GCE-NoGPU-x86_64-Debug-ASAN/builds/1155/steps/dm/logs/stdio

And causing crashes on Mac 10.6:
http://build.chromium.org/p/client.skia/builders/Test-Mac10.6-MacMini4.1-GeForce320M-x86_64-Debug/builds/1417/steps/dm/logs/stdio

Original issue's description:
> Fix Morphology effects sourcing outside of the crop rect.
>
> BUG=skia:1766
>
> Committed: https://skia.googlesource.com/skia/+/f6be925b5615f07039ce95c3433039694a8d1679

TBR=junov@google.com,junov@chromium.org,bsalomon@google.com,reed@google.com,cwallez@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:1766

Review URL: https://codereview.chromium.org/868973005

Cleanup: Delete buildbot_globals.py script.

rebaseline_server was the only user of this code but it was removed in
https://chromium.googlesource.com/skia/+/3886046f008f0542de36cbeb35e220404e7c88a9

So there aren't more users of this script and we can remove it now.

BUG=skia:3200
R=borenet@google.com

Review URL: https://codereview.chromium.org/873593005

Fix Morphology effects sourcing outside of the crop rect.

BUG=skia:1766

Review URL: https://codereview.chromium.org/781153002

Zap the scratch key on non-RT textures when scratch textures are disabled.

Review URL: https://codereview.chromium.org/870103002

Fix the speeling of "purgeable" in Gr code

TBR=robertphillips@google.com

Review URL: https://codereview.chromium.org/874693002

Cleanup SkPDFObject::emit*

Review URL: https://codereview.chromium.org/869783003

remove unused GM flags

Depends on https://codereview.chromium.org/873753002/

Thumbs up to CLion for refactoring this for me.

BUG=skia:

Review URL: https://codereview.chromium.org/867963004

Remove deprecated SkSurface::NewRenderTarget factories

Review URL: https://codereview.chromium.org/837723008

run clean branch baseline only once

BUG=skia:

Review URL: https://codereview.chromium.org/874643002

Spin off GM::runAsBench() from flags.

This will let us kill flags.

BUG=skia:

Review URL: https://codereview.chromium.org/873753002

Fix SampleApp picture-mode transform.

The multi-picture-draw path should not ignore the inherited transform.

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

Review URL: https://codereview.chromium.org/867303002

Take budgeted param when snapping new image.

Review URL: https://codereview.chromium.org/872543002

Fix 'Revert' instructions.

In the not so new pure Git world, 'svn fetch' and 'svn rebase' are not
needed anymore.

BUG=None
R=borenet@google.com

Review URL: https://codereview.chromium.org/872433004

Collect everything Skia into one document.

http://skiadocs.com:8000/dev/contrib/directory?cl=866133002

Review URL: https://codereview.chromium.org/866133002

check for too-large rowBytes

BUG=446164

Review URL: https://codereview.chromium.org/871993003