platform/upstream/libSkiaSharp.git
9 years agoUpdate {virtual,override} to follow C++11 style in src/pipe.
tfarina [Fri, 24 Apr 2015 20:12:44 +0000 (13:12 -0700)]
Update {virtual,override} to follow C++11 style in src/pipe.

The Google style guide states that only one of {virtual,override,final}
should be used for each declaration, since override implies virtual and
final implies both virtual and override.

The entries were found using the following command line:

$ find src/pipe -iname "*.h" -or -iname "*.cpp" | xargs pcregrep -M
"[^\n/]+virtual\ [^;{]+\ [a-zA-Z0-9_]+\([^;{]+\ override[ \n]*[;{]"

The regex was a courtesy of nick@chromium.org

BUG=None
R=mtklein@google.com

Review URL: https://codereview.chromium.org/1104863003

9 years agoDo not crash nanobench in debug mode
msarett [Fri, 24 Apr 2015 18:41:55 +0000 (11:41 -0700)]
Do not crash nanobench in debug mode

Instead print an error message

BUG=skia:

Review URL: https://codereview.chromium.org/1102083002

9 years agoFix some lingering typos
robertphillips [Fri, 24 Apr 2015 18:10:51 +0000 (11:10 -0700)]
Fix some lingering typos

Review URL: https://codereview.chromium.org/1101293003

9 years agoRevert of Remove validation exceptions in GrGLInterface. These are all now set by...
bsalomon [Fri, 24 Apr 2015 18:08:25 +0000 (11:08 -0700)]
Revert of Remove validation exceptions in GrGLInterface. These are all now set by Chromium. (patchset #1 id:1 of https://codereview.chromium.org/1100413003/)

Reason for revert:
Crashing bots

Original issue's description:
> Remove validation exceptions in GrGLInterface. These are all now set by Chromium.
>
> Committed: https://skia.googlesource.com/skia/+/42d9f35a8daf7f24ba542ba3783ec46fd965ef2f

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

Review URL: https://codereview.chromium.org/1093323005

9 years agosimple fix for preabandon valgrind error
joshualitt [Fri, 24 Apr 2015 16:57:12 +0000 (09:57 -0700)]
simple fix for preabandon valgrind error

TBR=bsalomon@google.com
BUG=skia:

Review URL: https://codereview.chromium.org/1106663003

9 years agofix multiple intersection logic
caryclark [Fri, 24 Apr 2015 16:08:57 +0000 (09:08 -0700)]
fix multiple intersection logic

When three or more curves intersect at the same point, ensure that
each curve records the intersections of the others. This fixes a
number of cubic tests.

TBR=reed@google.com
BUG=skia:3588

Review URL: https://codereview.chromium.org/1105943002

9 years agofix for using too much memory in GrBatchAtlas
joshualitt [Fri, 24 Apr 2015 15:33:21 +0000 (08:33 -0700)]
fix for using too much memory in GrBatchAtlas

BUG=skia:

Review URL: https://codereview.chromium.org/1105913002

9 years agoRemove validation exceptions in GrGLInterface. These are all now set by Chromium.
bsalomon [Fri, 24 Apr 2015 14:38:43 +0000 (07:38 -0700)]
Remove validation exceptions in GrGLInterface. These are all now set by Chromium.

Review URL: https://codereview.chromium.org/1100413003

9 years agoAdd ability to specify extra browser args and a prefix for SKP names to webpages_play...
rmistry [Thu, 23 Apr 2015 19:47:33 +0000 (12:47 -0700)]
Add ability to specify extra browser args and a prefix for SKP names to webpages_playback.py

BUG=skia:3763
NOTRY=true

Review URL: https://codereview.chromium.org/1066933006

9 years agoBlacklist desk_unicodetable SKP on Nexus Player
borenet [Thu, 23 Apr 2015 19:44:31 +0000 (12:44 -0700)]
Blacklist desk_unicodetable SKP on Nexus Player

It's getting us killed for OOM.

BUG=skia:3240

Review URL: https://codereview.chromium.org/1104693002

9 years agoRemove validation of blend barrier until Chrome can provide this function
bsalomon [Thu, 23 Apr 2015 19:25:51 +0000 (12:25 -0700)]
Remove validation of blend barrier until Chrome can provide this function
BUG=chromium:480534
TBR=egdaniel@google.com

Review URL: https://codereview.chromium.org/1101143002

9 years agoadd GM to show bad banding in raster for radial gradients
reed [Thu, 23 Apr 2015 17:30:27 +0000 (10:30 -0700)]
add GM to show bad banding in raster for radial gradients

BUG=427508
TBR=

Review URL: https://codereview.chromium.org/1061753009

9 years agoAdd onGetBlendInfo to GrXferProcessor
cdalton [Thu, 23 Apr 2015 16:40:23 +0000 (09:40 -0700)]
Add onGetBlendInfo to GrXferProcessor

Adds an onGetBlendInfo method for GrXferProcessor subclasses to
override instead of overriding getBlendInfo directly. This gives the
base class a chance to initialize the struct with default values
before passing it on. As the BlendInfo struct grows, this will keep
things simple and less error prone.

BUG=skia:
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/1049143002

9 years agoworking on initial winding for cubics
caryclark [Thu, 23 Apr 2015 16:13:37 +0000 (09:13 -0700)]
working on initial winding for cubics

Path ops works well for all tests except for cubics.
Isolate failures caused by cubics, and do a better job of computing
the initial winding for cubics.

TBR=reed@google.com
BUG=skia:3588

Review URL: https://codereview.chromium.org/1096923003

9 years agoEnsure that we create a NULL codec for images with zero dimensions
msarett [Thu, 23 Apr 2015 15:53:39 +0000 (08:53 -0700)]
Ensure that we create a NULL codec for images with zero dimensions

BUG=skia:3534
BUG=skia:3257

Review URL: https://codereview.chromium.org/1091043003

9 years agosmall fix for perf regression in GrAtlasTextContext.cpp
joshualitt [Thu, 23 Apr 2015 15:03:35 +0000 (08:03 -0700)]
small fix for perf regression in GrAtlasTextContext.cpp

BUG=skia:

Review URL: https://codereview.chromium.org/1080673005

9 years agouse acquire/release in SkEventTracer.cpp
mtklein [Thu, 23 Apr 2015 03:30:12 +0000 (20:30 -0700)]
use acquire/release in SkEventTracer.cpp

BUG=chromium:437044

Review URL: https://codereview.chromium.org/1091283006

9 years agoRemove unnecessary viewmatrix compare in GrAtlasTextContext
joshualitt [Thu, 23 Apr 2015 01:23:15 +0000 (18:23 -0700)]
Remove unnecessary viewmatrix compare in GrAtlasTextContext

BUG=skia:

Review URL: https://codereview.chromium.org/1088903009

9 years agonew GM to test clipping cubics
reed [Thu, 23 Apr 2015 00:38:23 +0000 (17:38 -0700)]
new GM to test clipping cubics

BUG=skia:3716
TBR=caryclark@google.com

Review URL: https://codereview.chromium.org/1096433009

9 years ago[SkPDFDevice] Enable pathops-based inverse fills
fmalita [Wed, 22 Apr 2015 22:02:03 +0000 (15:02 -0700)]
[SkPDFDevice] Enable pathops-based inverse fills

Keeping the old compile guard for clipping only.

BUG=skia:3749
R=halcanary@google.com,caryclark@google.com,reed@google.com

Review URL: https://codereview.chromium.org/1099953002

9 years agoRemove MSAA check from Ganesh veto (dashed paths are no longer slow).
senorblanco [Wed, 22 Apr 2015 21:33:34 +0000 (14:33 -0700)]
Remove MSAA check from Ganesh veto (dashed paths are no longer slow).

R=egdaniel@google.com
BUG=skia:

Review URL: https://codereview.chromium.org/1098753004

9 years agoMove function to be static instead of anonymous namespace
egdaniel [Wed, 22 Apr 2015 20:58:58 +0000 (13:58 -0700)]
Move function to be static instead of anonymous namespace

TBR=bsalomon@google.com, jvanverth@google.com

BUG=skia:

Review URL: https://codereview.chromium.org/1057363004

9 years agoFix for segfault on destruction of BitmapTextBlob
joshualitt [Wed, 22 Apr 2015 20:47:02 +0000 (13:47 -0700)]
Fix for segfault on destruction of BitmapTextBlob

BUG=skia:

Review URL: https://codereview.chromium.org/1065773005

9 years agoImprove the curve tessellation subdivision of the path renderers.
senorblanco [Wed, 22 Apr 2015 20:45:18 +0000 (13:45 -0700)]
Improve the curve tessellation subdivision of the path renderers.

The stencil-and-cover and tessellating path renderers currently
tessellate curves to linear segments to within one screen space pixel.
This is fairly poor accuracy, as can be seen from the strokecircle GM:
there are noticeable rotating bumps on the circles.

Improving the tolerance to 0.25 pixel eliminates the bumps, and
approximates the 16x supersampling of the raster path.
This does have a performance hit: 3-6% on desktop on
the IE chalkboard demo, ~1% on Nexus 7 2013.

Note: this will require rebaselining a number of GPU and  MSAA images
in Skia, but nothing in Chrome (yet).

BUG=skia:3731

Review URL: https://codereview.chromium.org/1072273007

9 years agoRefactor createTexture and onCreateTexture
egdaniel [Wed, 22 Apr 2015 20:27:39 +0000 (13:27 -0700)]
Refactor createTexture and onCreateTexture

BUG=skia:

Review URL: https://codereview.chromium.org/1102663002

9 years agoRevert of enable TextBlob / Batch (patchset #1 id:1 of https://codereview.chromium...
joshualitt [Wed, 22 Apr 2015 18:38:58 +0000 (11:38 -0700)]
Revert of enable TextBlob / Batch (patchset #1 id:1 of https://codereview.chromium.org/1077813004/)

Reason for revert:
seg fault

Original issue's description:
> enable TextBlob / Batch
>
> TBR=bsalomon@google.com
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/b75f8100ace48e74319ccf43d82deaac446784b2

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

Review URL: https://codereview.chromium.org/1096433007

9 years agoMake pixel_snap GMs incur less platform-specific variation
bsalomon [Wed, 22 Apr 2015 18:34:46 +0000 (11:34 -0700)]
Make pixel_snap GMs incur less platform-specific variation

Review URL: https://codereview.chromium.org/1080393009

9 years agoFix for GrAtlasTextContext memory leak
joshualitt [Wed, 22 Apr 2015 18:24:20 +0000 (11:24 -0700)]
Fix for GrAtlasTextContext memory leak

BUG=skia:

Review URL: https://codereview.chromium.org/1072193009

9 years agoRewrite and move Cluster Telemetry documentation to skia.org
rmistry [Wed, 22 Apr 2015 18:03:12 +0000 (11:03 -0700)]
Rewrite and move Cluster Telemetry documentation to skia.org

BUG=skia:3754
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=1074943003

Review URL: https://codereview.chromium.org/1074943003

9 years agoAdd tokens and entry points for KHR_blend_equation_advanced
cdalton [Wed, 22 Apr 2015 17:39:03 +0000 (10:39 -0700)]
Add tokens and entry points for KHR_blend_equation_advanced

Also adds glBlendEquation, which the extension interacts with, to the
core of GrGLInterface. Validation of this function is temporarily
disabled until Chrome hooks it up.

BUG=skia:

Review URL: https://codereview.chromium.org/1039693004

9 years agoAdd GM to test non-AA pixel snapping for points, lines, rects.
bsalomon [Wed, 22 Apr 2015 15:51:38 +0000 (08:51 -0700)]
Add GM to test non-AA pixel snapping for points, lines, rects.

Review URL: https://codereview.chromium.org/1070213003

9 years agoenable TextBlob / Batch
joshualitt [Wed, 22 Apr 2015 12:19:54 +0000 (05:19 -0700)]
enable TextBlob / Batch

TBR=bsalomon@google.com
BUG=skia:

Review URL: https://codereview.chromium.org/1077813004

9 years agoO(1) SkPictureUtils::ApproxBytesUsed()
mtklein [Tue, 21 Apr 2015 22:23:59 +0000 (15:23 -0700)]
O(1) SkPictureUtils::ApproxBytesUsed()

Chrome wants to call this more often, and it's quite slow today.

Seems like this could be clearer if SkPictureUtils::ApproxBytesUsed() were SkPicture::approxBytesUsed().

BUG=chromium:471873

Review URL: https://codereview.chromium.org/1090943004

9 years agoDefault to arm_v7_neon.
mtklein [Tue, 21 Apr 2015 21:25:34 +0000 (14:25 -0700)]
Default to arm_v7_neon.

This aliases all devices we know have NEON over to that too.

BUG=skia:

Review URL: https://codereview.chromium.org/1095953007

9 years agoRevert of Add tokens and entry points for KHR_blend_equation_advanced (patchset ...
tomhudson [Tue, 21 Apr 2015 21:16:53 +0000 (14:16 -0700)]
Revert of Add tokens and entry points for KHR_blend_equation_advanced (patchset #3 id:40001 of https://codereview.chromium.org/1039693004/)

Reason for revert:
We're getting repeated crashes since this CL landed when we try to evaluate a roll into Chrome on our bots. (https://uberchromegw.corp.google.com/i/client.skia/builders/Linux%20Tests/builds/1902)

Our initial reflex is to put a #if 0 around this code (just like the immediately following lines 334-338 in GrGLInterface.cpp), but the error isn't what the GPU team expected to see if that was all we needed, so we're reverting now and asking you to resubmit tomorrow after running more thorough tests.

Original issue's description:
> Add tokens and entry points for KHR_blend_equation_advanced
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/8e578859f80b46a63144add215955221017d3609

TBR=markkilgard@gmail.com,bsalomon@google.com,cdalton@nvidia.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/1101593002

9 years agofix some small bugs with GrAtlasTextContext
joshualitt [Tue, 21 Apr 2015 20:24:37 +0000 (13:24 -0700)]
fix some small bugs with GrAtlasTextContext

BUG=skia:

Review URL: https://codereview.chromium.org/1098093003

9 years agoAdd tokens and entry points for KHR_blend_equation_advanced
cdalton [Tue, 21 Apr 2015 19:34:09 +0000 (12:34 -0700)]
Add tokens and entry points for KHR_blend_equation_advanced

BUG=skia:

Review URL: https://codereview.chromium.org/1039693004

9 years agoAndroid's Clang toolchain can't assemble the JPEG's .S files.
mtklein [Tue, 21 Apr 2015 19:07:38 +0000 (12:07 -0700)]
Android's Clang toolchain can't assemble the JPEG's .S files.

Only affecting Build-Mac10.8-Clang-Arm7-Debug-Android.

BUG=skia:

Review URL: https://codereview.chromium.org/1093423003

9 years agoImport glTextureBarrier
cdalton [Tue, 21 Apr 2015 18:45:56 +0000 (11:45 -0700)]
Import glTextureBarrier

BUG=skia:

Review URL: https://codereview.chromium.org/1090163002

9 years agoSkip msaa4 on Tegra 2 and 3.
mtklein [Tue, 21 Apr 2015 18:25:47 +0000 (11:25 -0700)]
Skip msaa4 on Tegra 2 and 3.

This will fix the N7 and Xoom bots.

TBR=borenet@google.com

BUG=skia:

Review URL: https://codereview.chromium.org/1099943002

9 years agoRevert "One more ASAN fix."
senorblanco [Tue, 21 Apr 2015 18:13:17 +0000 (11:13 -0700)]
Revert "One more ASAN fix."

This reverts commit 409fd66a5afcef5f165f7ccec7c3473add231752.

We don't expect to pass this test until the ASAN bots are
correctly configured, and can run GPU tests.

BUG=skia:
TBR=robertphillips@google.com

Review URL: https://codereview.chromium.org/1080393003

9 years agoDM: failure to canary a sink is now fatal.
mtklein [Tue, 21 Apr 2015 17:49:13 +0000 (10:49 -0700)]
DM: failure to canary a sink is now fatal.

This requires we remove NVPR from the default set of configs, as we only find
out at runtime that it's not available.  All the other defaults will either be
compiled in and supported, or not compiled in and non-fatally skipped as
unknown configs.

BUG=skia:

Review URL: https://codereview.chromium.org/1100773003

9 years agoAsk for NVPR only when we expect it to be available.
mtklein [Tue, 21 Apr 2015 17:34:47 +0000 (10:34 -0700)]
Ask for NVPR only when we expect it to be available.

BUG=skia:

Review URL: https://codereview.chromium.org/1099023003

9 years agotypo in sherrifing docs
Greg Humphreys [Tue, 21 Apr 2015 17:28:44 +0000 (10:28 -0700)]
typo in sherrifing docs

BUG=skia:
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=1097353002

Review URL: https://codereview.chromium.org/1097353002

9 years agoSupport HWUI rename of DisplayListRenderer
tomhudson [Tue, 21 Apr 2015 17:09:07 +0000 (10:09 -0700)]
Support HWUI rename of DisplayListRenderer

R=djsollen@google.com,ccraik@google.com

Review URL: https://codereview.chromium.org/1086083004

9 years agoSome simple optimizations for improving GrAtlasTextContext perf
joshualitt [Tue, 21 Apr 2015 17:07:26 +0000 (10:07 -0700)]
Some simple optimizations for improving GrAtlasTextContext perf

BUG=skia:

Review URL: https://codereview.chromium.org/1093083003

9 years agoOne more ASAN fix.
senorblanco [Tue, 21 Apr 2015 16:43:08 +0000 (09:43 -0700)]
One more ASAN fix.

Don't fail the GL validation check if the GL context fails to create.

TBR=robertphillips@google.com
BUG=skia:

Review URL: https://codereview.chromium.org/1096113003

9 years agoAdd support for rotating / scaling / translating gpu cached distance field textblobs
joshualitt [Tue, 21 Apr 2015 16:43:03 +0000 (09:43 -0700)]
Add support for rotating / scaling / translating gpu cached distance field textblobs

BUG=skia:

Review URL: https://codereview.chromium.org/1092283002

9 years agoSplit default and user-supplied event tracers.
mtklein [Tue, 21 Apr 2015 16:38:03 +0000 (09:38 -0700)]
Split default and user-supplied event tracers.

Chrome's got a test scenario where setting the user-supplied tracer is racing with a concurrent use of the default tracer.

BUG=chromium:437044

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

Review URL: https://codereview.chromium.org/1099123002

9 years agoStore pointers to GrGlyph directly in BitmapTextBlob. This patch improves performanc...
joshualitt [Tue, 21 Apr 2015 16:37:57 +0000 (09:37 -0700)]
Store pointers to GrGlyph directly in BitmapTextBlob.  This patch improves performance by avoiding hashmap lookups under normal use

BUG=skia:

Review URL: https://codereview.chromium.org/1087203004

9 years agoLanding directly for testing the ios buildbot code.
stephana [Tue, 21 Apr 2015 15:31:59 +0000 (08:31 -0700)]
Landing directly for testing the ios buildbot code.

BUG=skia:
TBR=borenet

Review URL: https://codereview.chromium.org/1087653005

9 years agoRevert of Convert Color32 code to perfect blend. (patchset #6 id:100001 of https...
mtklein [Tue, 21 Apr 2015 15:09:30 +0000 (08:09 -0700)]
Revert of Convert Color32 code to perfect blend. (patchset #6 id:100001 of https://codereview.chromium.org/1098913002/)

Reason for revert:
Xfermode_SrcOver not looking encouraging.  Up to 50% regressions.

https://perf.skia.org/#3242

Original issue's description:
> Convert Color32 code to perfect blend.
>
> Before we commit to blend_256_round_alt, let's make sure blend_perfect is
> really slower in practice (i.e. regresses on perf.skia.org).
>
> blend_perfect is really the most desirable algorithm if we can afford it.  Not
> only is it correct, but it's easy to think about and break into correct pieces:
> for instance, its div255() doesn't require any coordination with the multiply.
>
> This looks like a 30% hit according to microbenches.  That said, microbenches
> said my previous change would be a 20-25% perf improvement, but it didn't end
> up showing a significant effect at a high level.
>
> As for correctness, I see a bunch of off-by-1 compared to blend_256_round_alt
> (exactly what we'd expect), and one off-by-3 in a GM that looks like it has a
> bunch of overdraw.
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/61221e7f87a99765b0e034020e06bb018e2a08c2

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

Review URL: https://codereview.chromium.org/1083923006

9 years agoPack BitmapTextBlob a bit tighter
joshualitt [Tue, 21 Apr 2015 15:03:04 +0000 (08:03 -0700)]
Pack BitmapTextBlob a bit tighter

BUG=skia:

Review URL: https://codereview.chromium.org/1092973004

9 years agoreuse scaler across consecutive textbatch flushes
joshualitt [Tue, 21 Apr 2015 14:49:49 +0000 (07:49 -0700)]
reuse scaler across consecutive textbatch flushes

BUG=skia:

Review URL: https://codereview.chromium.org/1091313002

9 years agoA small patch to enable distance field text in textblobs
joshualitt [Tue, 21 Apr 2015 14:35:10 +0000 (07:35 -0700)]
A small patch to enable distance field text in textblobs

BUG=skia:

Review URL: https://codereview.chromium.org/1098653005

9 years agoYet more ASAN fixes.
senorblanco [Tue, 21 Apr 2015 14:20:36 +0000 (07:20 -0700)]
Yet more ASAN fixes.

More nullptr checks for factories I have added.
Other checks more Yoda-like I have made. (Skia style this is.)

BUG=skia:

Review URL: https://codereview.chromium.org/1086393004

9 years agoAnother speculative ASAN fix.
senorblanco [Tue, 21 Apr 2015 13:59:17 +0000 (06:59 -0700)]
Another speculative ASAN fix.

Add a GrContext nullptr check in SurfaceTest.

TBR=robertphillips@google.com

Review URL: https://codereview.chromium.org/1098993004

9 years agoSkTHash: remove()
mtklein [Tue, 21 Apr 2015 13:53:56 +0000 (06:53 -0700)]
SkTHash: remove()

BUG=skia:

Review URL: https://codereview.chromium.org/1057043003

9 years agoUndo the bounds mapping of bounds for GrOvalRenderer for certain draws
egdaniel [Tue, 21 Apr 2015 13:28:08 +0000 (06:28 -0700)]
Undo the bounds mapping of bounds for GrOvalRenderer for certain draws

BUG=skia:

Review URL: https://codereview.chromium.org/1099873002

9 years agofix valgrind uninitialized var error
caryclark [Tue, 21 Apr 2015 12:27:02 +0000 (05:27 -0700)]
fix valgrind uninitialized var error

R=mtklein@google.com

Review URL: https://codereview.chromium.org/1097293002

9 years agoSpeculative fix for ASAN build.
senorblanco [Mon, 20 Apr 2015 22:00:06 +0000 (15:00 -0700)]
Speculative fix for ASAN build.

Check for NULL GrContext in TessellatingPathRendererTests.
Also fix an upload nag (no #includes before GPU #ifdef).

TBR=bsalomon@google.com

Review URL: https://codereview.chromium.org/1097943004

9 years agoMake the GPU dashing effect MSAA-friendly.
senorblanco [Mon, 20 Apr 2015 21:44:26 +0000 (14:44 -0700)]
Make the GPU dashing effect MSAA-friendly.

Review URL: https://codereview.chromium.org/1092793006

9 years agoRegularize lmp::XXX_element_handler.
bungeman [Mon, 20 Apr 2015 20:31:46 +0000 (13:31 -0700)]
Regularize lmp::XXX_element_handler.

This makes all non-XMLCALL element_handler functions uniform by making
their signature

void XXX_element_handler(FamilyData* self, const char** attributes)

This makes the code somewhat easier to follow, as it makes explicit
which functions are actually logically methods of FamilyData (start
with 'FamilyData* self') and which are called directly by expat
(start with 'void* data').

Review URL: https://codereview.chromium.org/1086943007

9 years agoMinor modifications to GrProgramDesc.h.
jvanverth [Mon, 20 Apr 2015 19:29:37 +0000 (12:29 -0700)]
Minor modifications to GrProgramDesc.h.

BUG=skia:

Review URL: https://codereview.chromium.org/1100483003

9 years agoReturn correctly mapped bounds from GrOvalRenderer.
egdaniel [Mon, 20 Apr 2015 19:28:57 +0000 (12:28 -0700)]
Return correctly mapped bounds from GrOvalRenderer.

Before we were returning non transformed bounds for the ovals instead of
device bounds from the various draw calls. This was causing us to do dst
copies from the wrong portion of the screen when there was some transform
on the matrix.

This fixes the gpu issues from the new imagefilters-xfermodes gm as well an
old bug from the srcmode GM.

BUG=skia:

Review URL: https://codereview.chromium.org/1090453003

9 years agoRegularize informative messages in Android font parser.
bungeman [Mon, 20 Apr 2015 18:16:48 +0000 (11:16 -0700)]
Regularize informative messages in Android font parser.

The Android font parser prints a number of helpful messages when
something seems amiss in the configuration file. Give all of these
messages a common and more informative format.

Review URL: https://codereview.chromium.org/1096063003

9 years agoConvert Color32 code to perfect blend.
mtklein [Mon, 20 Apr 2015 17:52:26 +0000 (10:52 -0700)]
Convert Color32 code to perfect blend.

Before we commit to blend_256_round_alt, let's make sure blend_perfect is
really slower in practice (i.e. regresses on perf.skia.org).

blend_perfect is really the most desirable algorithm if we can afford it.  Not
only is it correct, but it's easy to think about and break into correct pieces:
for instance, its div255() doesn't require any coordination with the multiply.

This looks like a 30% hit according to microbenches.  That said, microbenches
said my previous change would be a 20-25% perf improvement, but it didn't end
up showing a significant effect at a high level.

As for correctness, I see a bunch of off-by-1 compared to blend_256_round_alt
(exactly what we'd expect), and one off-by-3 in a GM that looks like it has a
bunch of overdraw.

BUG=skia:

Review URL: https://codereview.chromium.org/1098913002

9 years agoadjust GM to have a final opaque bg, to make it easier to triage
reed [Mon, 20 Apr 2015 16:10:31 +0000 (09:10 -0700)]
adjust GM to have a final opaque bg, to make it easier to triage

BUG=skia:
TBR=

Review URL: https://codereview.chromium.org/1085333003

9 years agoNow, path ops natively intersect conics, quads, and cubics in any combination. There...
caryclark [Mon, 20 Apr 2015 15:31:59 +0000 (08:31 -0700)]
Now, path ops natively intersect conics, quads, and cubics in any combination. There are still a class of cubic tests that fail and a handful of undiagnosed failures from skps and fuzz tests, but things are much better overall.

Extended tests (150M+) run to completion in release in about 6 minutes; the standard test suite exceeds 100K and finishes in a few seconds on desktops.

TBR=reed
BUG=skia:3588

Review URL: https://codereview.chromium.org/1037953004

9 years agoapply xfermode after imagefilter when we create a tmp layer
reed [Mon, 20 Apr 2015 15:04:21 +0000 (08:04 -0700)]
apply xfermode after imagefilter when we create a tmp layer

BUG=skia:3741

Review URL: https://codereview.chromium.org/1091173003

9 years agoadd GM to test imagefilters + xfermodes
reed [Mon, 20 Apr 2015 14:27:15 +0000 (07:27 -0700)]
add GM to test imagefilters + xfermodes

draws incorrectly as is, but will be fixed by https://codereview.chromium.org/1091173003

BUG=skia:3741

Review URL: https://codereview.chromium.org/1088773004

9 years agoUpdate our libjpeg to the version used by Chromium.
scroggo [Mon, 20 Apr 2015 13:58:49 +0000 (06:58 -0700)]
Update our libjpeg to the version used by Chromium.

We had previously not updated since 2013. This includes the fix
for http://seclists.org/fulldisclosure/2013/Nov/83

BUG=skia:3744

Review URL: https://codereview.chromium.org/1097943002

9 years agoFix point count computation in tessellating path renderer.
senorblanco [Mon, 20 Apr 2015 12:41:48 +0000 (05:41 -0700)]
Fix point count computation in tessellating path renderer.

The conic, quadratic and cubic computations were using tolerance^2
instead of tolerance when computing maximum point count, causing paths
to be undertessellated when magnifying and overtessellated when minifying.

(Funny story: this bug went unnoticed back when we were tessellating
paths in screen space, when tolerance and its square were both 1.)

BUG=skia:3731

Review URL: https://codereview.chromium.org/1095773003

9 years agoremove dead code from SK_SUPPORT_LEGACY_BLITANTIH2V2
reed [Mon, 20 Apr 2015 12:18:01 +0000 (05:18 -0700)]
remove dead code from SK_SUPPORT_LEGACY_BLITANTIH2V2

no image changes expected in skia or chrome/blink

BUG=skia:
TBR=

Review URL: https://codereview.chromium.org/1050953003

9 years agoUpdate SKP version
skia.buildbots [Sun, 19 Apr 2015 08:15:08 +0000 (01:15 -0700)]
Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=

Review URL: https://codereview.chromium.org/1095953002

9 years agoRevert of Allow NEON on iOS. (patchset #3 id:40001 of https://codereview.chromium...
mtklein [Fri, 17 Apr 2015 21:02:40 +0000 (14:02 -0700)]
Revert of Allow NEON on iOS. (patchset #3 id:40001 of https://codereview.chromium.org/1091823002/)

Reason for revert:
need to coordinate this with chrome gyps/gn

Original issue's description:
> Allow NEON on iOS.
>
> I have nanobench building and running (Color32_arm_neon) on my iPad.
>
> No public API changes.
> TBR=reed@google.com
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/e5043b7ea5170a639367b67c986568907576be4b

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

Review URL: https://codereview.chromium.org/1094843005

9 years agoRespect declared font style on Android.
bungeman [Fri, 17 Apr 2015 20:51:08 +0000 (13:51 -0700)]
Respect declared font style on Android.

Previously the normal/italic style bit was obtained from scanning the
font file. With the new format the style may be stated explicitly, and
this explicit value in the configuration file should override any
information obtained from the font data itself.

This change allows the font element's style attribute to override the
font's style, but retains the default 'auto' setting for backwards
compatibility. Repecting the style bit may become more important with
variation fonts, because it will be up to the configuration writer to
determine what values of the 'slnt' variation should be considered
'normal' or 'italic'.

DOCS_PREVIEW= https://skia.org/?cl=1092093002

Committed: https://skia.googlesource.com/skia/+/673e902c9b9982a167f54f1cc175d8d9cab8bcaf

Review URL: https://codereview.chromium.org/1092093002

9 years agoRevert of Respect declared font style on Android. (patchset #6 id:100001 of https...
jcgregorio [Fri, 17 Apr 2015 20:30:35 +0000 (13:30 -0700)]
Revert of Respect declared font style on Android. (patchset #6 id:100001 of https://codereview.chromium.org/1092093002/)

Reason for revert:
Failed on the compile bots.

Original issue's description:
> Respect declared font style on Android.
>
> Previously the normal/italic style bit was obtained from scanning the
> font file. With the new format the style may be stated explicitly, and
> this explicit value in the configuration file should override any
> information obtained from the font data itself.
>
> This change allows the font element's style attribute to override the
> font's style, but retains the default 'auto' setting for backwards
> compatibility. Repecting the style bit may become more important with
> variation fonts, because it will be up to the configuration writer to
> determine what values of the 'slnt' variation should be considered
> 'normal' or 'italic'.
>
> DOCS_PREVIEW= https://skia.org/?cl=1092093002
>
> Committed: https://skia.googlesource.com/skia/+/673e902c9b9982a167f54f1cc175d8d9cab8bcaf

TBR=mtklein@google.com,tomhudson@google.com,scroggo@google.com,bungeman@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/1082173004

9 years agoPDF: Correctly embed JPEG images directly into PDF output.
halcanary [Fri, 17 Apr 2015 20:27:24 +0000 (13:27 -0700)]
PDF: Correctly embed JPEG images directly into PDF output.

We only embed images with YUV planes.  That should only grab the
subset of color JPEGs supported by PDF.

BUG=skia:3180

Review URL: https://codereview.chromium.org/1025773002

9 years agoRespect declared font style on Android.
bungeman [Fri, 17 Apr 2015 20:25:03 +0000 (13:25 -0700)]
Respect declared font style on Android.

Previously the normal/italic style bit was obtained from scanning the
font file. With the new format the style may be stated explicitly, and
this explicit value in the configuration file should override any
information obtained from the font data itself.

This change allows the font element's style attribute to override the
font's style, but retains the default 'auto' setting for backwards
compatibility. Repecting the style bit may become more important with
variation fonts, because it will be up to the configuration writer to
determine what values of the 'slnt' variation should be considered
'normal' or 'italic'.

DOCS_PREVIEW= https://skia.org/?cl=1092093002

Review URL: https://codereview.chromium.org/1092093002

9 years agoTrace contention for lockPixels mutex
enne [Fri, 17 Apr 2015 19:42:49 +0000 (12:42 -0700)]
Trace contention for lockPixels mutex

Review URL: https://codereview.chromium.org/1090103002

9 years agoRework SSE and NEON Color32 algorithms to be more correct and faster.
mtklein [Fri, 17 Apr 2015 18:00:54 +0000 (11:00 -0700)]
Rework SSE and NEON Color32 algorithms to be more correct and faster.

This algorithm changes the blend math, guarded by SK_LEGACY_COLOR32_MATH.  The new math is more correct: it's never off by more than 1, and correct in all the interesting 0x00 and 0xFF edge cases, where the old math was never off by more than 2, and not always correct on the edges.

If you look at tests/BlendTest.cpp, the old code was using the `blend_256_plus1_trunc` algorithm, while the new code uses `blend_256_round_alt`.  Neither uses `blend_perfect`, which is about ~35% slower than `blend_256_round_alt`.

This will require an unfathomable number of rebaselines, first to Skia, then to Blink when I remove the guard.

I plan to follow up with some integer SIMD abstractions that can unify these two implementations into a single algorithm.  This was originally what I was working on here, but the correctness gains seem to be quite compelling.  The only places these two algorithms really differ greatly now is the kernel function, and even there they can really both be expressed abstractly as:
  - multiply 8-bits and 8-bits producing 16-bits
  - add 16-bits to 16-bits, returning the top 8 bits.
All the constants are the same, except SSE is a little faster to keep 8 16-bit inverse alphas, NEON's a little faster to keep 8 8-bit inverse alphas.  I may need to take this small speed win back to unify the two.

We should expect a ~25% speedup on Intel (mostly from unrolling to 8 pixels) and a ~20% speedup on ARM (mostly from using vaddhn to add `color`, round, and narrow back down to 8-bit all into one instruction.

(I am probably missing several more related bugs here.)
BUG=skia:3738,skia:420,chromium:111470

Review URL: https://codereview.chromium.org/1092433002

9 years agoRemove filename from SkFontDescriptor.
bungeman [Fri, 17 Apr 2015 18:00:06 +0000 (11:00 -0700)]
Remove filename from SkFontDescriptor.

No one actually uses it, so remove it. Keep the string id for backward
compatibility.

TBR=reed@google.com
This doesn't change any public API.

Review URL: https://codereview.chromium.org/1057413005

9 years agoSmall change to allow DistanceField and BMP text to coexist in a blob
joshualitt [Fri, 17 Apr 2015 17:32:32 +0000 (10:32 -0700)]
Small change to allow DistanceField and BMP text to coexist in a blob

BUG=skia:

Review URL: https://codereview.chromium.org/1096753002

9 years agoCorrect SkFixedToDouble.
bungeman [Fri, 17 Apr 2015 17:05:43 +0000 (10:05 -0700)]
Correct SkFixedToDouble.

SkFixedToDouble uses the decimal floating point string representation
of 1/(2^16) to provide a constant value. This value is exactly
representable by a double but the decimal constant currently used
does not map to this value.

This change modifies this constant to more decimal digits so that
the actual double value desired is created.

Review URL: https://codereview.chromium.org/1097473002

9 years agoInitial CL to add distance field support to GrAtlasTextContext
joshualitt [Fri, 17 Apr 2015 16:30:06 +0000 (09:30 -0700)]
Initial CL to add distance field support to GrAtlasTextContext

BUG=skia:

Review URL: https://codereview.chromium.org/1082843002

9 years agoadding new gm to rotate, translate, and scale textblobs
joshualitt [Fri, 17 Apr 2015 16:01:05 +0000 (09:01 -0700)]
adding new gm to rotate, translate, and scale textblobs

this is to test distance field caching and regen of texture blobs on scale.

BUG=skia:

Review URL: https://codereview.chromium.org/1097563002

9 years agoAllow NEON on iOS.
mtklein [Fri, 17 Apr 2015 13:36:53 +0000 (06:36 -0700)]
Allow NEON on iOS.

I have nanobench building and running (Color32_arm_neon) on my iPad.

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

BUG=skia:

Review URL: https://codereview.chromium.org/1091823002

9 years agoadd a GM to demo the blend bug
mtklein [Fri, 17 Apr 2015 13:32:13 +0000 (06:32 -0700)]
add a GM to demo the blend bug

BUG=skia:3739

Review URL: https://codereview.chromium.org/1094863003

9 years agoRevert of Revert of remove unused (by clients) SkPathUtils (patchset #1 id:1 of https...
reed [Fri, 17 Apr 2015 03:10:55 +0000 (20:10 -0700)]
Revert of Revert of remove unused (by clients) SkPathUtils (patchset #1 id:1 of https://codereview.chromium.org/1060703003/)

Reason for revert:
fix (removal from gypi/gn files) has landed in chrome.

Original issue's description:
> Revert of remove unused (by clients) SkPathUtils (patchset #1 id:1 of https://codereview.chromium.org/1088383003/)
>
> Reason for revert:
> This change is causing the DEPS roll to fail:
>
>
> http://build.chromium.org/p/tryserver.chromium.linux/builders/android_chromium_gn_compile_rel/builds/78771/steps/gn/logs/stdio
>
> Original issue's description:
> > remove unused (by clients) SkPathUtils
> >
> > BUG=skia:
> >
> > Committed: https://skia.googlesource.com/skia/+/aab35d91b8b80acd1902594bbf542083fdfa4bb7
>
> TBR=scroggo@google.com,reed@chromium.org
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/bdb0bf5f8858043878d8a4fa8130c6c87bef3fd4

TBR=scroggo@google.com,jcgregorio@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/1091963002

9 years agoAdd PS 'all' weight name to scanFont.
bungeman [Thu, 16 Apr 2015 21:22:46 +0000 (14:22 -0700)]
Add PS 'all' weight name to scanFont.

Multiple Master fonts often list their weight as 'all'. However, they
usually default to normal weight. State explicitly that we know about
this weight string.

Review URL: https://codereview.chromium.org/1092793002

9 years agoDefault iOS builds to our code signing identity.
mtklein [Thu, 16 Apr 2015 20:29:58 +0000 (13:29 -0700)]
Default iOS builds to our code signing identity.

It's super tedious to keep setting it every time I rerun ./gyp_skia.
Doesn't look like it'll bother the bots.

BUG=skia:

Review URL: https://codereview.chromium.org/1053493005

9 years agocontrib: Simplify the command to create a feature branch.
tfarina [Thu, 16 Apr 2015 20:28:06 +0000 (13:28 -0700)]
contrib: Simplify the command to create a feature branch.

You can create a work branch off origin/master by simply running:

$ git checkout -b my-feature origin/master

BUG=None
R=borenet@google.com
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=1057513004

Review URL: https://codereview.chromium.org/1057513004

9 years agoThis section is moot: we divert ios to opts_none above.
mtklein [Thu, 16 Apr 2015 19:45:38 +0000 (12:45 -0700)]
This section is moot: we divert ios to opts_none above.

(We never set arm_version = 7 for iOS...)

CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Mac10.7-Clang-Arm7-Debug-iOS-Trybot

BUG=skia:

Review URL: https://codereview.chromium.org/1092753002

9 years agoAdd SkTPin.
bungeman [Thu, 16 Apr 2015 19:18:28 +0000 (12:18 -0700)]
Add SkTPin.

Currently there exist SkScalarPin and SkPin32, and in a future change
another version is desired. This change introduces SkTPin and changes
SkScalarPin and SkPin32 to use it.

Review URL: https://codereview.chromium.org/1090003002

9 years agoImprove tessellating path rasterizer performance on dashed paths.
senorblanco [Thu, 16 Apr 2015 18:47:18 +0000 (11:47 -0700)]
Improve tessellating path rasterizer performance on dashed paths.

With dashed paths, we get a lot of geometry on the same sort line.
Since the vertices are sorted, it makes more sense to insert edges
from the end (in sort order) than the beginning. This requires
maintaining both the head and tail of the active edge list.

This gives a ~20% perf improvement on tabl_digg.skp.

BUG=skia:3733

Review URL: https://codereview.chromium.org/1089903002

9 years agoadd textblob cache freeall
joshualitt [Thu, 16 Apr 2015 18:24:04 +0000 (11:24 -0700)]
add textblob cache freeall

BUG=skia:

Review URL: https://codereview.chromium.org/1088683004

9 years agoRename GrStencilBuffer to GrStencilAttachment
egdaniel [Thu, 16 Apr 2015 18:22:42 +0000 (11:22 -0700)]
Rename GrStencilBuffer to GrStencilAttachment

BUG=skia:

Review URL: https://codereview.chromium.org/1083133002

9 years agoSkScaledBitmapSampler: fix memory overwritten
zoran.jovanovic [Thu, 16 Apr 2015 18:03:16 +0000 (11:03 -0700)]
SkScaledBitmapSampler: fix memory overwritten

Memory will be overwritten while downsampling some
interlaced gif images, most commonly with odd sizes,
when index of destination row stores in the current
line computed from GifInterlaceIter meets:

 X is an integer in [0..height-1]
   and
 (X < height)
 && ((X - sampleSize/2) % sampleSize == 0)
 && ((X - sampleSize/2)/sampleSize >= height/sampleSize)

Signed-off-by: Lu Tong <lu.x.tong@sonymobile.com>
BUG=skia:

Review URL: https://codereview.chromium.org/1085253002

9 years agoAdd ReverseIter to GrTRecorder
cdalton [Thu, 16 Apr 2015 17:42:49 +0000 (10:42 -0700)]
Add ReverseIter to GrTRecorder

BUG=skia:

Review URL: https://codereview.chromium.org/1035083004