On gpu, use max(r,g,b) for coverage alpha in LCD and update lcd blend gm.

BUG=skia:

Review URL: https://codereview.chromium.org/1339213004

Update params to get better packing for DF path atlas

Gets about a 4x speedup on desk_chalkboard.skp.

Review URL: https://codereview.chromium.org/1345823002

Forward declare SkStrokeRec in SkPathEffect

Review URL: https://codereview.chromium.org/1312163008

create SkBitmapProvider to abstract images and bitmaps

BUG=skia:

Review URL: https://codereview.chromium.org/1340223003

Convert unit tests, GMs from SkBitmapSource to SkImagesource

This removes SkBitmapSource clients within Skia.

http://crrev.com/1334173004 does the same for Blink, so we should be
able to remove SkBitmapSource in a follow-up.

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

Review URL: https://codereview.chromium.org/1343123002

Add special case circle blur for Ganesh

This makes the blurcircles bench go from ~33us to ~8us on Windows desktop.

It will require layout test suppressions

Review URL: https://codereview.chromium.org/1311583005

small tweaks to cmake_build

  - streamline how we define $here
  - only make `cmake` when bootstrapping; it's a little less to build

BUG=skia:4269
CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Mac10.9-Clang-x86_64-Release-CMake-Trybot,Build-Ubuntu-GCC-x86_64-Release-CMake-Trybot

Review URL: https://codereview.chromium.org/1345663004

Add support for blending of LCD for all blend modes.
BUG=skia:

Review URL: https://codereview.chromium.org/1313623002

Minor fix to attaching stencils

BUG=skia:

Review URL: https://codereview.chromium.org/1344033002

Add viewer mode to VisualBench.

Displays each benchmark/skp with a graph showing a series of frame times.
Use the space bar to skip to the next benchmark.

Adds an option to hit ESC to quit VisualBench. Useful in fullscreen mode.

Review URL: https://codereview.chromium.org/1336043003

Parallel cache.

TBR=reed@google.com

BUG=skia:1330,528560

Committed: https://skia.googlesource.com/skia/+/6f2a486040cb25465990196c229feb47e668e87f

Committed: https://skia.googlesource.com/skia/+/bf2988833e5a36c6b430da6fdd2cfebd0015adec

Review URL: https://codereview.chromium.org/1264103003

Add documentation of cs.skia.org.

BUG=skia:
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=1332303005

Review URL: https://codereview.chromium.org/1332303005

Speculative fix for uninitialized bitmap mem in ImageFilterTest

TBR=reed@google.com

Review URL: https://codereview.chromium.org/1341923003

Let's try building the example app on the bots too.

BUG=skia:4269

CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Mac10.9-Clang-x86_64-Release-CMake-Trybot,Build-Ubuntu-GCC-x86_64-Release-CMake-Trybot
NOTREECHECKS=true

Review URL: https://codereview.chromium.org/1344843002

cmake_build is a bash script.

BUG=skia:4269

Review URL: https://codereview.chromium.org/1334203003 .

Suppress test_imagegenerator_factory

Temporarily disable the test while we figure what (if anything) to do
about the SkImageGenerator gFactory race.

BUG=skia:4339
TBR=reed@google.com,mtklein@google.com

Review URL: https://codereview.chromium.org/1344693002

remove code from SK_SUPPORT_LEGACY_NEWFROMGENERATOR, eliminates caller of deprecated SkInstallDiscardablePixelRef

BUG=skia:

Review URL: https://codereview.chromium.org/1344663002

fix leaked generator

BUG=skia:4335

Review URL: https://codereview.chromium.org/1336353003

SkImageSource

Blink is migrating away from SkBitmaps, so we need an SkImage-based
SkImageFilter source.  This is pretty much a 1-1 equivalent of
SkBitmapSource.

To avoid duplication, relocate the SkImage deserialization logic
from SkPictureData to SkReadBuffer.

R=reed@google.com,robertphillips@google.com,senorblanco@chromium.org

Review URL: https://codereview.chromium.org/1343703005

Add helper for creating leaf FPs inside GrFP::TestCreate functions

Review URL: https://codereview.chromium.org/1334273003

Revert of SkPx: new approach to fixed-point SIMD (patchset #9 id:160001 of https://codereview.chromium.org/1317233005/ )

Reason for revert:
http://build.chromium.org/p/client.skia.compile/builders/Build-Mac10.8-Clang-Arm7-Debug-Android/builds/4627

Original issue's description:
> SkPx: new approach to fixed-point SIMD
>
> SkPx is like Sk4px, except each platform implementation of SkPx can declare
> a different sweet spot of N pixels, with extra loads and stores to handle the
> ragged edge of 0<n<N pixels.
>
> In this case, _sse's sweet spot remains 4 pixels.   _neon jumps up to 8 so
> we can now use NEON's transposing loads and stores, and _none is just 1.
> This makes operations involving alpha considerably more efficient on NEON,
> as alpha is its own distinct 8x8 bit plane that's easy to toss around.
>
> This incorporates a few other improvements I've been wanting:
>   - no requirement that we're dealing with SkPMColor.  SkColor works too.
>   - no anonymous namespace hack to differentiate implementations.
>
> Codegen and perf look good on Clang/x86-64 and GCC/ARMv7.
> The NEON code looks very similar to the old NEON code, as intended.
> No .skp or GM diffs on my laptop.  Don't expect any.
>
> I intend this to replace Sk4px.  Plan after landing:
>   - port SkXfermode_opts.h
>   - port Color32 in SkBlitRow_D32.cpp (and move to SkBlitRow_opts.h like other
>     SkOpts code)
>   - delete all Sk4px-related code
>   - clean up evolutionary dead ends in SkNx (Sk16b, Sk16h, Sk4i, Sk4d, etc.)
>     leaving Sk2f, Sk4f (and Sk2s, Sk4s).
>   - find a machine with AVX2 to work on, write SkPx_avx2.h handling 8 pixels
>     at a time.
>
> In the end we'll have Sk4f for float pixels, SkPx for fixed-point pixels.
>
> BUG=skia:4117
>
> Committed: https://skia.googlesource.com/skia/+/82c93b45ed6ac0b628adb8375389c202d1f586f9

TBR=mtklein@google.com,msarett@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:4117

Review URL: https://codereview.chromium.org/1336423002

Move some of the adding stencil attachment logic of Gpu and into Render Target.

The new flow of calls for attaching a Stencil looks like:

Client
  rt->attachStencilAttachment()
    gpu->getStencilAttachment()
      glgpu->createStencilAttachment()
    glrt->completeStencilAttachment() //actually attaches

BUG=skia:

Review URL: https://codereview.chromium.org/1333383002

Update SkWhitelistChecksums.cpp with the checksums of the fonts on the CT slave machines.

This file was generated using the whitelist_typefaces tool.
I tried out this file on 3 of the 100
CT slaves and './out/Debug/whitelist_typefaces --check' successfully returned 0.

BUG=skia:4336

Review URL: https://codereview.chromium.org/1344553004

SkPx: new approach to fixed-point SIMD

SkPx is like Sk4px, except each platform implementation of SkPx can declare
a different sweet spot of N pixels, with extra loads and stores to handle the
ragged edge of 0<n<N pixels.

In this case, _sse's sweet spot remains 4 pixels.   _neon jumps up to 8 so
we can now use NEON's transposing loads and stores, and _none is just 1.
This makes operations involving alpha considerably more efficient on NEON,
as alpha is its own distinct 8x8 bit plane that's easy to toss around.

This incorporates a few other improvements I've been wanting:
  - no requirement that we're dealing with SkPMColor.  SkColor works too.
  - no anonymous namespace hack to differentiate implementations.

Codegen and perf look good on Clang/x86-64 and GCC/ARMv7.
The NEON code looks very similar to the old NEON code, as intended.
No .skp or GM diffs on my laptop.  Don't expect any.

I intend this to replace Sk4px.  Plan after landing:
  - port SkXfermode_opts.h
  - port Color32 in SkBlitRow_D32.cpp (and move to SkBlitRow_opts.h like other
    SkOpts code)
  - delete all Sk4px-related code
  - clean up evolutionary dead ends in SkNx (Sk16b, Sk16h, Sk4i, Sk4d, etc.)
    leaving Sk2f, Sk4f (and Sk2s, Sk4s).
  - find a machine with AVX2 to work on, write SkPx_avx2.h handling 8 pixels
    at a time.

In the end we'll have Sk4f for float pixels, SkPx for fixed-point pixels.

BUG=skia:4117

Review URL: https://codereview.chromium.org/1317233005

Test that GrFragmentProcessors work without input colors.

Committed: https://skia.googlesource.com/skia/+/72c58e7052af2a0855412ce4b249f977069db751

Review URL: https://codereview.chromium.org/1341853002

cmake_build: support SKIA_OUT and BUILDTYPE

And, fix BUILDTYPE=Debug build.
EQUAL is for numbers, STREQUAL for strings.

BUG=skia:

Review URL: https://codereview.chromium.org/1341763003

Revert of Test that GrFragmentProcessors work without input colors. (patchset #2 id:20001 of https://codereview.chromium.org/1341853002/ )

Reason for revert:
Need to fix up more processor subclasses.

Original issue's description:
> Test that GrFragmentProcessors work without input colors.
>
> Committed: https://skia.googlesource.com/skia/+/72c58e7052af2a0855412ce4b249f977069db751

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

Review URL: https://codereview.chromium.org/1338403003

Test that GrFragmentProcessors work without input colors.

Review URL: https://codereview.chromium.org/1341853002

Fix GPU-only snapping bug in mask blur rendering

The existing mask effect code in Ganesh is subject to snapping issues (when the created mask is redrawn). This artifact can be seen by rendering the original geometry (w/o blurs) and comparing that result to a rendering of the unblurred masks. W/o this patch the results do not match up (they are arbitrarily shifted by a pixel).

This patch will require rebaselining and suppressions.

Chromium layout tests suppressions are here:
https://codereview.chromium.org/1342683003/ (Add layout test suppressions for upcoming Skia roll)

Review URL: https://codereview.chromium.org/1338183002

impl preroll for all image backends

BUG=skia:

Review URL: https://codereview.chromium.org/1341043002

Have SkVarAlloc::alloc() use sk_malloc_throw.

Very right, it's not prepared to handle return-NULL mallocs at all.

BUG=530759

Review URL: https://codereview.chromium.org/1339093002

SkValidatingReadBuffer: make sure we don't call readTypeface().

Add an assert to make sure we're not calling this unimplemented method.

BUG=skia:3978

Review URL: https://codereview.chromium.org/1341753003

fix leak in test

BUG=skia:4335

Review URL: https://codereview.chromium.org/1336763007

be sure to use cached bitmap when we need to upload something to make a texture

BUG=skia:4334

Review URL: https://codereview.chromium.org/1338373002

CMake bot script

 - Add CMake v3.3.1 (latest) to DEPS.
 - Add cmake/bot-cmake.sh to bootstrap CMake then build Skia using that.

Works on my Mac and Linux box, both with no system CMake installation.

CMake will be ~100M on disk.  The first bootstrap takes a couple minutes,
and a no-op re-run of bot-cmake.sh takes 15-20 seconds.  I thought about
having bot-cmake.sh fetch CMake instead of DEPS, but I'm not sure I can
handle updates, etc. as robustly as it can.

This will only work on Linux and Mac.  CMake requires an older CMake on
Windows.  It doesn't have an equivalent ./bootstrap there.  Will have to
think about how Windows bots will work!

BUG=skia:4269

Review URL: https://codereview.chromium.org/1339603003

we must own/free the generator, even if we fail to return a cacherator

BUG=skia:4332

Review URL: https://codereview.chromium.org/1345523002

disable kIndex_8 gpu support for now -- seems to always be slower

BUG=skia:4333

Review URL: https://codereview.chromium.org/1339103002

discardable pixelrefs are gone, update tests accordingly

BUG=skia:4328

Review URL: https://codereview.chromium.org/1340803002

remove dead code not mentioned in any GYP

BUG=skia:

Review URL: https://codereview.chromium.org/1340563003

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=
NO_MERGE_BUILDS

Review URL: https://codereview.chromium.org/1338933002

formalize generate->bitmap

just move block of code to expose it

BUG=skia:4328
TBR=

Review URL: https://codereview.chromium.org/1334033004

remove obsolete samples

BUG=skia:
TBR=

Review URL: https://codereview.chromium.org/1340793002

sync-and-gyp: Update shell script to correct the syntax of functions

R=halcanary@gmail.com

Review URL: https://codereview.chromium.org/1342543003 .

skia: Add ANGLE with GL backend to nanobench/DM

This will allow us to test this without hacking it in, might be useful
for others too.

Review URL: https://codereview.chromium.org/1338003002

move GrGLPathProcessor back into GrPathProcessor

BUG=skia:

Review URL: https://codereview.chromium.org/1333423003

support colortables in cacherator

BUG=skia:
TBR=scroggo

Review URL: https://codereview.chromium.org/1339753002

remove path specific program building classes

BUG=skia:

Review URL: https://codereview.chromium.org/1336763003

Provide link to worker logs.

BUG=skia:4063
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=1338823002

Review URL: https://codereview.chromium.org/1338823002

Start trying to collapse path program stuff

BUG=skia:

Review URL: https://codereview.chromium.org/1333273003

Remove jpegs with uninitialized memory from Gold

BitmapRegionSampler (uses SkImageDecoder) will often scale
to a power of 2 regardless of the sampleSize requested.
This is skbug.com/4319.

Consider a 60x60 image.
To decode a subset with sample size 3, we might ask for the
following.
[x, y, w, h] = [-15, -15, 30, 30]
sampleSize = 3

Since w = 30 and h = 30, this should give us a 10x10
result.  Only the bottom right 5x5 quadrant of this 10x10
subset will actually be in the image.  We should get a 5
pixel border on the top and left because we ask for 15
extra pixels on the top and left.

Unfortunately, SkImageDecoder will take our requested
sample size of 3, and then decide to use a sample size of
2.  Not only will it scale the image by 2, but it will also
scale the border by 2.  So while we are expecting pixel
data to begin at offset (5, 5) of the result bitmap, it
actually begins at offset (7, 7).  Making things worse,
the pixels between (5, 5) and (7, 7) are uninitialized,
causing problems on Gold.

Options for fixing this include:
(1) Not testing decodes with a border.
(2) Changing the test to check the size of the output
bitmap.
(3) Disable the tests.

I think it's best to just disable these tests.  We know
they don't work, so why do we need to see the results on
Gold?

BUG=skia:4319

Review URL: https://codereview.chromium.org/1313233007

Remove batchtracker

BUG=skia:

Review URL: https://codereview.chromium.org/1332923003

increase resource image-cache size to test perf

BUG=skia:
TBR=

Review URL: https://codereview.chromium.org/1332393002

Revert of Parallel cache - preliminary  (patchset #23 id:440001 of https://codereview.chromium.org/1264103003/ )
Also reverts https://codereview.chromium.org/1333003002/ which was layered on top.

Reason for revert:
Appears to leak GDI handles: http://build.chromium.org/p/chromium.memory.fyi/builders/Windows%20Unit%20%28DrMemory%20full%29%20%282%29/builds/8247

~~Dr.M~~ Error #1: HANDLE LEAK: GDI handle 0x03050a84 and 3 similar handle(s) were opened but not closed:
~~Dr.M~~ # 0 system call NtGdiCreateDIBSection
~~Dr.M~~ # 1 GDI32.dll!CreateDIBSection                                                +0xdc     (0x768ead23 <GDI32.dll+0x1ad23>)
~~Dr.M~~ # 2 skia.dll!HDCOffscreen::draw                                                [third_party\skia\src\ports\skfonthost_win.cpp:499]
~~Dr.M~~ # 3 skia.dll!SkScalerContext_GDI::generateImage                                [third_party\skia\src\ports\skfonthost_win.cpp:1233]
~~Dr.M~~ # 4 skia.dll!SkScalerContext::getImage                                         [third_party\skia\src\core\skscalercontext.cpp:530]
~~Dr.M~~ # 5 skia.dll!SkGlyphCache::OnceFillInImage                                     [third_party\skia\src\core\skglyphcache.cpp:252]
~~Dr.M~~ # 6 skia.dll!sk_once_slow<>                                                    [third_party\skia\include\core\skonce.h:76]
~~Dr.M~~ # 7 skia.dll!SkGlyphCache::findImage                                           [third_party\skia\src\core\skglyphcache.cpp:260]
~~Dr.M~~ # 8 skia.dll!D1G_RectClip                                                      [third_party\skia\src\core\skdraw.cpp:1479]
~~Dr.M~~ # 9 skia.dll!SkDraw::drawPosText                                               [third_party\skia\src\core\skdraw.cpp:1838]
~~Dr.M~~ #10 skia.dll!SkBitmapDevice::drawPosText                                       [third_party\skia\src\core\skbitmapdevice.cpp:348]
~~Dr.M~~ #11 skia.dll!SkCanvas::onDrawPosText                                           [third_party\skia\src\core\skcanvas.cpp:2433]
~~Dr.M~~ #12 skia.dll!SkCanvas::drawPosText                                             [third_party\skia\src\core\skcanvas.cpp:2507]
~~Dr.M~~ #13 skia.dll!SkRecords::Draw::draw<>                                           [third_party\skia\src\core\skrecorddraw.cpp:109]
~~Dr.M~~ #14 skia.dll!SkRecord::Record::visit<>                                         [third_party\skia\src\core\skrecord.h:170]
~~Dr.M~~ #15 skia.dll!SkRecordDraw                                                      [third_party\skia\src\core\skrecorddraw.cpp:55]
~~Dr.M~~ #16 skia.dll!SkBigPicture::playback                                            [third_party\skia\src\core\skbigpicture.cpp:43]
~~Dr.M~~ #17 skia.dll!SkCanvas::onDrawPicture                                           [third_party\skia\src\core\skcanvas.cpp:2800]
~~Dr.M~~ #18 skia.dll!SkCanvas::drawPicture                                             [third_party\skia\src\core\skcanvas.cpp:2770]
~~Dr.M~~ #19 cc.dll!cc::DrawingDisplayItem::Raster                                      [cc\playback\drawing_display_item.cc:51]
~~Dr.M~~ #20 cc.dll!cc::DisplayItemList::Raster                                         [cc\playback\display_item_list.cc:107]
~~Dr.M~~ #21 cc.dll!cc::DisplayListRasterSource::RasterCommon                           [cc\playback\display_list_raster_source.cc:122]
~~Dr.M~~ #22 cc.dll!cc::DisplayListRasterSource::PlaybackToCanvas                       [cc\playback\display_list_raster_source.cc:100]
~~Dr.M~~ #23 cc.dll!cc::TileTaskWorkerPool::PlaybackToMemory                            [cc\raster\tile_task_worker_pool.cc:208]
~~Dr.M~~ #24 cc.dll!cc::OneCopyTileTaskWorkerPool::PlaybackAndCopyOnWorkerThread        [cc\raster\one_copy_tile_task_worker_pool.cc:413]
~~Dr.M~~ #25 cc.dll!cc::`anonymous namespace'::RasterBufferImpl::Playback               [cc\raster\one_copy_tile_task_worker_pool.cc:53]
~~Dr.M~~ #26 cc.dll!cc::`anonymous namespace'::RasterTaskImpl::Raster                   [cc\tiles\tile_manager.cc:131]
~~Dr.M~~ #27 cc.dll!cc::`anonymous namespace'::RasterTaskImpl::RunOnWorkerThread        [cc\tiles\tile_manager.cc:90]
~~Dr.M~~ #28 cc.dll!cc::TaskGraphRunner::RunTaskWithLockAcquired                        [cc\raster\task_graph_runner.cc:418]
~~Dr.M~~ #29 cc.dll!cc::TaskGraphRunner::Run                                            [cc\raster\task_graph_runner.cc:361]
~~Dr.M~~ #30 base.dll!base::SimpleThread::ThreadMain                                    [base\threading\simple_thread.cc:66]
~~Dr.M~~ #31 base.dll!base::`anonymous namespace'::ThreadFunc                           [base\threading\platform_thread_win.cc:82]
~~Dr.M~~ #32 KERNEL32.dll!BaseThreadInitThunk                                          +0x11     (0x7570337a <KERNEL32.dll+0x1337a>)
~~Dr.M~~ Note: @0:15:51.087 in thread 196
~~Dr.M~~ Note: handles created with the same callstack are closed here:
~~Dr.M~~ Note: # 0 system call NtGdiDeleteObjectApp
~~Dr.M~~ Note: # 1 GDI32.dll!DeleteObject                                                    +0x149    (0x768e57d3 <GDI32.dll+0x157d3>)
~~Dr.M~~ Note: # 2 skia.dll!HDCOffscreen::draw                                                [third_party\skia\src\ports\skfonthost_win.cpp:471]
~~Dr.M~~ Note: # 3 skia.dll!SkScalerContext_GDI::generateImage                                [third_party\skia\src\ports\skfonthost_win.cpp:1233]
~~Dr.M~~ Note: # 4 skia.dll!SkScalerContext::getImage                                         [third_party\skia\src\core\skscalercontext.cpp:530]
~~Dr.M~~ Note: # 5 skia.dll!SkGlyphCache::OnceFillInImage                                     [third_party\skia\src\core\skglyphcache.cpp:252]
~~Dr.M~~ Note: # 6 skia.dll!sk_once_slow<>                                                    [third_party\skia\include\core\skonce.h:76]
~~Dr.M~~ Note: # 7 skia.dll!SkGlyphCache::findImage                                           [third_party\skia\src\core\skglyphcache.cpp:260]
~~Dr.M~~ Note: # 8 skia.dll!D1G_RectClip                                                      [third_party\skia\src\core\skdraw.cpp:1479]
~~Dr.M~~ Note: # 9 skia.dll!SkDraw::drawPosText                                               [third_party\skia\src\core\skdraw.cpp:1838]
~~Dr.M~~ Note: #10 skia.dll!SkBitmapDevice::drawPosText                                       [third_party\skia\src\core\skbitmapdevice.cpp:348]
~~Dr.M~~ Note: #11 skia.dll!SkCanvas::onDrawPosText                                           [third_party\skia\src\core\skcanvas.cpp:2433]
~~Dr.M~~ Note: #12 skia.dll!SkCanvas::drawPosText                                             [third_party\skia\src\core\skcanvas.cpp:2507]
~~Dr.M~~ Note: #13 skia.dll!SkRecords::Draw::draw<>                                           [third_party\skia\src\core\skrecorddraw.cpp:109]
~~Dr.M~~ Note: #14 skia.dll!SkRecord::Record::visit<>                                         [third_party\skia\src\core\skrecord.h:170]
~~Dr.M~~ Note: #15 skia.dll!SkRecordDraw                                                      [third_party\skia\src\core\skrecorddraw.cpp:55]
~~Dr.M~~ Note: #16 skia.dll!SkBigPicture::playback                                            [third_party\skia\src\core\skbigpicture.cpp:43]
~~Dr.M~~ Note: #17 skia.dll!SkCanvas::onDrawPicture                                           [third_party\skia\src\core\skcanvas.cpp:2800]
~~Dr.M~~ Note: #18 skia.dll!SkCanvas::drawPicture                                             [third_party\skia\src\core\skcanvas.cpp:2770]
~~Dr.M~~ Note: #19 cc.dll!cc::DrawingDisplayItem::Raster                                      [cc\playback\drawing_display_item.cc:51]
~~Dr.M~~ Note: #20 cc.dll!cc::DisplayItemList::Raster                                         [cc\playback\display_item_list.cc:107]
~~Dr.M~~ Note: #21 cc.dll!cc::DisplayListRasterSource::RasterCommon                           [cc\playback\display_list_raster_source.cc:122]
~~Dr.M~~ Note: #22 cc.dll!cc::DisplayListRasterSource::PlaybackToCanvas                       [cc\playback\display_list_raster_source.cc:100]
~~Dr.M~~ Note: #23 cc.dll!cc::TileTaskWorkerPool::PlaybackToMemory                            [cc\raster\tile_task_worker_pool.cc:208]
~~Dr.M~~ Note: #24 cc.dll!cc::OneCopyTileTaskWorkerPool::PlaybackAndCopyOnWorkerThread        [cc\raster\one_copy_tile_task_worker_pool.cc:413]
~~Dr.M~~ Note: #25 cc.dll!cc::`anonymous namespace'::RasterBufferImpl::Playback               [cc\raster\one_copy_tile_task_worker_pool.cc:53]
~~Dr.M~~ Note: #26 cc.dll!cc::`anonymous namespace'::RasterTaskImpl::Raster                   [cc\tiles\tile_manager.cc:131]
~~Dr.M~~ Note: #27 cc.dll!cc::`anonymous namespace'::RasterTaskImpl::RunOnWorkerThread        [cc\tiles\tile_manager.cc:90]
~~Dr.M~~ Note: #28 cc.dll!cc::TaskGraphRunner::RunTaskWithLockAcquired                        [cc\raster\task_graph_runner.cc:418]
~~Dr.M~~ Note: #29 cc.dll!cc::TaskGraphRunner::Run                                            [cc\raster\task_graph_runner.cc:361]
~~Dr.M~~ Note: #30 base.dll!base::SimpleThread::ThreadMain                                    [base\threading\simple_thread.cc:66]
~~Dr.M~~ Note: #31 base.dll!base::`anonymous namespace'::ThreadFunc                           [base\threading\platform_thread_win.cc:82]
~~Dr.M~~ Note: #32 KERNEL32.dll!BaseThreadInitThunk                                          +0x11     (0x7570337a <KERNEL32.dll+0x1337a>)

Original issue's description:
> Parallel cache.
>
> TBR=reed@google.com
>
> BUG=skia:1330
>
> Committed: https://skia.googlesource.com/skia/+/6f2a486040cb25465990196c229feb47e668e87f
>
> Committed: https://skia.googlesource.com/skia/+/bf2988833e5a36c6b430da6fdd2cfebd0015adec

TBR=reed@google.com,mtklein@google.com,mtklein@chromium.org,herb@google.com
BUG=skia:1330

[mtklein mucking around]
NOTREECHECKS=true

Review URL: https://codereview.chromium.org/1339493002

Revert of use new shuffle to speed up affine matrix mappts (patchset #3 id:40001 of https://codereview.chromium.org/1333983002/ )

Reason for revert:
Unexpected perf impact, and a whole bunch of new images in gold (mostly invisibly different).

Original issue's description:
> use new shuffle to speed up affine matrix mappts
>
> sse: 25 -> 18
> neon: 95 -> 86
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/e70afc9f48d00828ee6b707899a8ff542b0e8b98

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

Review URL: https://codereview.chromium.org/1335003002

Use SkImageCacherator in SkImages

Possible follow-up changes to consider

1. Roll SkImage_Raster and _Gpu into _Generator, where the generator (or cacherator) is backed by a pre-existing texture or raster.
2. Evolve SkImageUsageType into a verb requiring stretching, and have the caller (common code) digest the caps() and usage, so that subclasses are just told what to do (stretch or not)
3. Common code/utility to convert an unstretched texture into a stretch one (and cache it) if the generator can only make an unstretched one.

BUG=skia:

Review URL: https://codereview.chromium.org/1282363002

use new shuffle to speed up affine matrix mappts

sse: 25 -> 18
neon: 95 -> 86

BUG=skia:

Review URL: https://codereview.chromium.org/1333983002

Add simd.md to document Skia SIMD code / plans.

BUG=skia:4117
NOTRY=true
DOCS_PREVIEW= https://skia.org/dev/contrib/simd?cl=1330083002

Review URL: https://codereview.chromium.org/1330083002

Correct a possible free after use.

This shows up in TSAN runs as a use-after-free warning.

BUG=skia:

Review URL: https://codereview.chromium.org/1333003002

SkNx_shuffle

This allows us to express shuffles more directly in code while also giving us a
convenient point to platform-specify particular shuffles for particular types.

No specializations yet.  Everyone just uses the (pretty good) default option.

BUG=skia:

Review URL: https://codereview.chromium.org/1301413006

fix missing clipmaskmanager

Review URL: https://codereview.chromium.org/1330213004 .

make a shallow-copy so we don't get racy trying to lock/unlock our private bitmap

BUG= 529995
NOTREECHECKS=True

Review URL: https://codereview.chromium.org/1327313002

Fix texture creation on stencil format test code

BUG=skia:

Review URL: https://codereview.chromium.org/1332853003

Port SkMatrix opts to SkOpts.

No changes to the code, just moved around.

This will have the effect of enabling vectorized code on ARMv7.
Should be no effect on ARMv8 or x86, which would have been vectorized already.

nanobench --match mappoints changes on Nexus 5 (ARMv7):

_affine: 132 -> 95
_scale: 118 -> 47
_trans: 60 -> 37

A teaser:
We should next look at the ABCD->BADC shuffle we've noted that we need in _affine.  A quick hack showed doing that optimally is another ~35% speedup on x86.  Got to figure out how to do it best on ARM though: that same quick hack was a 2x slowdown there.  Good reason to resurrect that SkNx_shuffle() CL!

(I believe the answers are vrev64q_f32(v) and _mm_shuffle_ps(v,v, _MM_SHUFFLE(2,3,0,1), but we should probably find out in another CL.)

BUG=skia:4117

Review URL: https://codereview.chromium.org/1320673014

Remove GrClipTarget

Review URL: https://codereview.chromium.org/1330353006

Simplify installation of pipeling into GrDrawBatch in GrDrawTarget

R=joshualitt@google.com

Review URL: https://codereview.chromium.org/1333943002

Late creation of GrPathProcessor

BUG=skia:

Review URL: https://codereview.chromium.org/1337513002

Cleanup GrDrawTarget now that all paths lead to GrBatch

Review URL: https://codereview.chromium.org/1315513008

Port SkBlitRow::Color32 to SkOpts.

This was a pre-SkOpts attempt that we can bring under its wing now.

This should be a perf no-op, deo volente.

BUG=skia:4117

Review URL: https://codereview.chromium.org/1314863006

DM: add --key and --properties to PNG description field.

This tags each image with the bot that created it.

BUG=skia:4322

Review URL: https://codereview.chromium.org/1332903002

Fix leak of path ranges

R=egdaniel@google.com

Review URL: https://codereview.chromium.org/1334753002

Calculate pixel config and stencil fmt pairs once per pixel config.

We use a temp FB and stencil buffer to test different stencil formats with
a given pixel config. We then keep a map from pixel config to desired stencil
format.

BUG=skia:

Review URL: https://codereview.chromium.org/1317443004

Add a mutex to GrContext::readSurfacePixels to protect against multiple CPU raster threads accessing the same GrContext to read back GPU input data

BUG=chromium:524717

TBR=reed@google.com

Committed: https://skia.googlesource.com/skia/+/eb662bc407cec0585a821946fef123102cae64db

Review URL: https://codereview.chromium.org/1329313002

There's a set probability that a linear pipeline of random procs will be created (old behavior), or a pipeline with a single proc tree (added behavior).

Had to move GrComposeEffect class definition from SkComposeShader.cpp to SkComposeShader.h so that GLProgramsTest can call GrComposeEffect::Create()

BUG=skia:4182

Review URL: https://codereview.chromium.org/1314923002

Initialize subrun variables to make Valgrind happy

The check at line 1259 was failing in Valgrind, and in
the default constructor subrun.fMaskFormat is never inited.
Also set some other variables to avoid future problems.

Review URL: https://codereview.chromium.org/1315773005

Revert of Add a mutex to GrContext::readSurfacePixels to protect against multiple CPU raster threads accessin… (patchset #1 id:1 of https://codereview.chromium.org/1329313002/ )

Reason for revert:
breaking the bots

Original issue's description:
> Add a mutex to GrContext::readSurfacePixels to protect against multiple CPU raster threads accessing the same GrContext to read back GPU input data
>
> BUG=chromium:524717
>
> TBR=reed@google.com
>
> Committed: https://skia.googlesource.com/skia/+/eb662bc407cec0585a821946fef123102cae64db

TBR=reed@google.com,reed@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:524717

Review URL: https://codereview.chromium.org/1334603002

Add a mutex to GrContext::readSurfacePixels to protect against multiple CPU raster threads accessing the same GrContext to read back GPU input data

BUG=chromium:524717

TBR=reed@google.com

Review URL: https://codereview.chromium.org/1329313002

SK_API on GrWrapTextureInBitmap

BUG=chromium:524717
TBR=reed@google.com

Review URL: https://codereview.chromium.org/1334563002

Remove race.

BUG=skia:

Review URL: https://codereview.chromium.org/1327193003

fix leaked picture

BUG=skia:
TBR=

Review URL: https://codereview.chromium.org/1329143006

don't peek w/o first acquiring the lock

BUG=skia:

Review URL: https://codereview.chromium.org/1324503005

Reland of more zero-length changes for svg compatibility (patchset #1 id:1 of https://codereview.chromium.org/1334543002/ )

Reason for revert:
DEPS should be fixed now

Original issue's description:
> Revert of more zero-length changes for svg compatibility (patchset #6 id:100001 of https://codereview.chromium.org/1330623003/ )
>
> Reason for revert:
> breaks DEPS
>
> Original issue's description:
> > more zero-length changes for svg compatibility
> >
> > If a path contains a moveTo followed by a line or curve,
> > even if the line or curve has no length, SVG expects
> > the end caps to draw if the cap style is round or square.
> >
> > Fredrik Söderquist attached a patch to the chrome bug
> > (slightly modified here) that fixes layout test failures
> > resulting from deleting special-case code in SVG
> > dealing with zero-length path segments.
> >
> > R=reed@google.com,fs@opera.com
> > BUG=22974
> >
> > Committed: https://skia.googlesource.com/skia/+/62fb1ba1786863e545c89839b5706ad5151cec15
>
> TBR=fs@opera.com,reed@google.com
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=22974
>
> Committed: https://skia.googlesource.com/skia/+/5ca4fa3846067a47e88d35ace895df3ebe3ec2a5

TBR=fs@opera.com,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=22974

Review URL: https://codereview.chromium.org/1314833004

Brian requested a new namespace for factories that create fragment processors that combine 2 things with a xfermode.  This way, we can change the implementation of these factories in the future to not use GrComposeEffect if desired.

BUG=skia:4182

Review URL: https://codereview.chromium.org/1306163005

Parallel cache.

TBR=reed@google.com

BUG=skia:1330

Committed: https://skia.googlesource.com/skia/+/6f2a486040cb25465990196c229feb47e668e87f

Review URL: https://codereview.chromium.org/1264103003

add picture-image variant

BUG=skia:

Review URL: https://codereview.chromium.org/1329283002

Mimic Chrome better in our GYPs.

To match Chrome, make sure we've disabled thread-safe statics, RTTI, and exceptions.  Linux needed -fno-threadsafe-statics, Mac needed all three.

Nothing important triggered this CL.  I just got confused when I saw exception handling (calls to delete, stack unwinding) in some generated code on my laptop.

BUG=skia:

Review URL: https://codereview.chromium.org/1323533003

switch over to using images

BUG=skia:
TBR=

Review URL: https://codereview.chromium.org/1313543004

Specialize SkOncePtr<T[]>.

SkOncePtr<T[]> is identical to SkOncePtr<T> except we'll default to delete[]
for cleanup.

CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Debug-ASAN-Trybot

BUG=skia:

No public API changes.
TBR=reed@google.com

Review URL: https://codereview.chromium.org/1311893010

Make GrProcessorDataManager a noop

TBR=bsalomon@google.com
BUG=skia:

Review URL: https://codereview.chromium.org/1323963003

GrPathRangeBatch

BUG=skia:

Review URL: https://codereview.chromium.org/1315563003

Revert of more zero-length changes for svg compatibility (patchset #6 id:100001 of https://codereview.chromium.org/1330623003/ )

Reason for revert:
breaks DEPS

Original issue's description:
> more zero-length changes for svg compatibility
>
> If a path contains a moveTo followed by a line or curve,
> even if the line or curve has no length, SVG expects
> the end caps to draw if the cap style is round or square.
>
> Fredrik Söderquist attached a patch to the chrome bug
> (slightly modified here) that fixes layout test failures
> resulting from deleting special-case code in SVG
> dealing with zero-length path segments.
>
> R=reed@google.com,fs@opera.com
> BUG=22974
>
> Committed: https://skia.googlesource.com/skia/+/62fb1ba1786863e545c89839b5706ad5151cec15

TBR=fs@opera.com,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=22974

Review URL: https://codereview.chromium.org/1334543002

Port uses of SkLazyPtr to SkOncePtr.

This gives SkOncePtr a non-trivial destructor that uses std::default_delete
by default.  This is overrideable, as seen in SkColorTable.

SK_DECLARE_STATIC_ONCE_PTR still just leaves its pointers hanging at EOP.

BUG=skia:

No public API changes.
TBR=reed@google.com

Committed: https://skia.googlesource.com/skia/+/a1254acdb344174e761f5061c820559dab64a74c

Review URL: https://codereview.chromium.org/1322933005

wire up incNumDraws

BUG=skia:

Review URL: https://codereview.chromium.org/1333623003

GM: replace boilerplate with macros

I have verified locally that nothing draws differently.

Motivation:
*   SK_SIMPLE_GM makes it easier to write a GM.
*   Reducing 1100 lines of code makes maintenance easier.
*   Simple GMs are easy to convert to Fiddles.

Review URL: https://codereview.chromium.org/1333553002

Detach cache in DF routines rather than passing it in.

This removes nesting of cache accesses, which avoids a possible
deadlock condition when the caches are the same.

Review URL: https://codereview.chromium.org/1310513008

Make SkGraphics::Term a no-op, stop calling it.

I'd remove it entirely but Android is calling it explicitly.

BUG=skia:4259

Committed: https://skia.googlesource.com/skia/+/925979f733fe8e70d84627147dee04d030423349

Review URL: https://codereview.chromium.org/1329853005

Revert of Make SkGraphics::Term a no-op, stop calling it. (patchset #2 id:20001 of https://codereview.chromium.org/1329853005/ )

Reason for revert:
SK_ATTR_DEPRECATED is meaningful to Android.  Don't use it.

Original issue's description:
> Make SkGraphics::Term a no-op, stop calling it.
>
> I'd remove it entirely but Android is calling it explicitly.
>
> BUG=skia:4259
>
> Committed: https://skia.googlesource.com/skia/+/925979f733fe8e70d84627147dee04d030423349

TBR=reed@google.com,scroggo@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:4259

Review URL: https://codereview.chromium.org/1314483006

Revert of Port uses of SkLazyPtr to SkOncePtr. (patchset #7 id:110001 of https://codereview.chromium.org/1322933005/ )

Reason for revert:
Breaks Chrome roll.

obj/skia/ext/skia_chrome.skia_memory_dump_provider.o
does not have -I include/private on its include path, but transitively includes SkMessageBus.h.

Original issue's description:
> Port uses of SkLazyPtr to SkOncePtr.
>
> This gives SkOncePtr a non-trivial destructor that uses std::default_delete
> by default.  This is overrideable, as seen in SkColorTable.
>
> SK_DECLARE_STATIC_ONCE_PTR still just leaves its pointers hanging at EOP.
>
> BUG=skia:
>
> No public API changes.
> TBR=reed@google.com
>
> Committed: https://skia.googlesource.com/skia/+/a1254acdb344174e761f5061c820559dab64a74c

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

Review URL: https://codereview.chromium.org/1334523002

more zero-length changes for svg compatibility

If a path contains a moveTo followed by a line or curve,
even if the line or curve has no length, SVG expects
the end caps to draw if the cap style is round or square.

Fredrik Söderquist attached a patch to the chrome bug
(slightly modified here) that fixes layout test failures
resulting from deleting special-case code in SVG
dealing with zero-length path segments.

R=reed@google.com,fs@opera.com
BUG=22974

Review URL: https://codereview.chromium.org/1330623003

Port uses of SkLazyPtr to SkOncePtr.

This gives SkOncePtr a non-trivial destructor that uses std::default_delete
by default.  This is overrideable, as seen in SkColorTable.

SK_DECLARE_STATIC_ONCE_PTR still just leaves its pointers hanging at EOP.

BUG=skia:

No public API changes.
TBR=reed@google.com

Review URL: https://codereview.chromium.org/1322933005

Provides various implementations of Android's SkBitmapRegionDecoder.

Implements testing in DM for these implementations.

nanobench testing will follow after this.

TBR=scroggo
BUG=skia:

Committed: https://skia.googlesource.com/skia/+/76f755e6d54a32f9887ad254ce59a3a62f28bde4

Review URL: https://codereview.chromium.org/1288963002