Aarch64 support
This change contains the necessary modifications to have Skia build and
run properly on an ARMv8 processor in aarch64 execution state.
Here's a list of the changes:
- add an arm64 target to the build system + SK_CPU_ARM64 flag
- MatrixTest was failing when built in Release mode. Fused MAC
instructions were generated which made some intermediate results
more accurate. As the test relies on result comparison, the more
precise results when compared to others led to a gap bigger than
what was tolerated. As I don't know if some actual skia code relies
on results being comparable, I've disabled fused MAC instruction
with -ffp-contract=off for arm64.
- Modify include/core/SkOnce.h to have barriers work.
- SK_CPU_ARM64 implies SK_ARM_NEON_MODE_ALWAYS.
- use existing Xfermode optimisations with modifications that can be
removed in the future when toolchains are ready. Also save a few
instructions is two Xfermodes (will apply to ARM too).
- use existing SkBoxBlur and SkMorphology optimisations.
- use existing SkBlitMask optimisations
- use existing BitmapProcState and Convolution optimisations.
Future changes will include:
- Blitters (only partialy merged upstream)
- SkUtils (there's little value in sending asm optimisations without
having them benchmarked on real hardware).
Signed-off-by: Kevin PETIT <kevin.petit@arm.com>
BUG=skia:
Committed: http://code.google.com/p/skia/source/detail?r=13980R=djsollen@google.com, reed@google.com, mtklein@google.com, halcanary@google.com
Author: kevin.petit@arm.com
Review URL: https://codereview.chromium.org/143423004
git-svn-id: http://skia.googlecode.com/svn/trunk@14025 2bbb7eff-a529-9590-31e7-b0007b416f81
If one of inputs to SkXfermodeImageFilter draws nothing, either due to
it being cropped out upstream, or within the filter itself, the filter
should still draw the other input, since otherwise the result will be incorrect.
For the GPU path, since we can't detect this case in
canFilterImageGPU() without recursing, we'll just drop to
the generic path if either input is empty, since we can't use the effect in that case anyway.
While we're at it, let's drop to the generic path if the
xfermode can't be expressed as an effect, since the code
here was doing a 2-pass render in that case anyway, which
is equivalent to what the (xfermode == NULL) case was doing
anyway.
R=bsalomon@google.com, sugoi@chromium.org
Review URL: https://codereview.chromium.org/220723007
git-svn-id: http://skia.googlecode.com/svn/trunk@14016 2bbb7eff-a529-9590-31e7-b0007b416f81
Reason for revert:
GYP's failing on most (all?) bots.
Original issue's description:
> ARM Skia NEON patches - 35 - First AArch64 support
>
> Aarch64 support
>
> This change contains the necessary modifications to have Skia build and
> run properly on an ARMv8 processor in aarch64 execution state.
>
> Here's a list of the changes:
>
> - add an arm64 target to the build system + SK_CPU_ARM64 flag
>
> - MatrixTest was failing when built in Release mode. Fused MAC
> instructions were generated which made some intermediate results
> more accurate. As the test relies on result comparison, the more
> precise results when compared to others led to a gap bigger than
> what was tolerated. As I don't know if some actual skia code relies
> on results being comparable, I've disabled fused MAC instruction
> with -ffp-contract=off for arm64.
>
> - Modify include/core/SkOnce.h to have barriers work.
>
> - SK_CPU_ARM64 implies SK_ARM_NEON_MODE_ALWAYS.
>
> - use existing Xfermode optimisations with modifications that can be
> removed in the future when toolchains are ready. Also save a few
> instructions is two Xfermodes (will apply to ARM too).
>
> - use existing SkBoxBlur and SkMorphology optimisations.
>
> - use existing SkBlitMask optimisations
>
> - use existing BitmapProcState and Convolution optimisations.
>
> Future changes will include:
>
> - Blitters (only partialy merged upstream)
>
> - SkUtils (there's little value in sending asm optimisations without
> having them benchmarked on real hardware).
>
> Signed-off-by: Kevin PETIT <kevin.petit@arm.com>
>
> BUG=skia:
>
> Committed: http://code.google.com/p/skia/source/detail?r=13980R=djsollen@google.com, reed@google.com, halcanary@google.com, kevin.petit@arm.comTBR=djsollen@google.com, halcanary@google.com, kevin.petit@arm.com, reed@google.com
NOTREECHECKS=true
NOTRY=true
BUG=skia:
Author: mtklein@google.com
Review URL: https://codereview.chromium.org/216113005
git-svn-id: http://skia.googlecode.com/svn/trunk@13983 2bbb7eff-a529-9590-31e7-b0007b416f81
Aarch64 support
This change contains the necessary modifications to have Skia build and
run properly on an ARMv8 processor in aarch64 execution state.
Here's a list of the changes:
- add an arm64 target to the build system + SK_CPU_ARM64 flag
- MatrixTest was failing when built in Release mode. Fused MAC
instructions were generated which made some intermediate results
more accurate. As the test relies on result comparison, the more
precise results when compared to others led to a gap bigger than
what was tolerated. As I don't know if some actual skia code relies
on results being comparable, I've disabled fused MAC instruction
with -ffp-contract=off for arm64.
- Modify include/core/SkOnce.h to have barriers work.
- SK_CPU_ARM64 implies SK_ARM_NEON_MODE_ALWAYS.
- use existing Xfermode optimisations with modifications that can be
removed in the future when toolchains are ready. Also save a few
instructions is two Xfermodes (will apply to ARM too).
- use existing SkBoxBlur and SkMorphology optimisations.
- use existing SkBlitMask optimisations
- use existing BitmapProcState and Convolution optimisations.
Future changes will include:
- Blitters (only partialy merged upstream)
- SkUtils (there's little value in sending asm optimisations without
having them benchmarked on real hardware).
Signed-off-by: Kevin PETIT <kevin.petit@arm.com>
BUG=skia:
R=djsollen@google.com, reed@google.com, mtklein@google.com, halcanary@google.com
Author: kevin.petit@arm.com
Review URL: https://codereview.chromium.org/143423004
git-svn-id: http://skia.googlecode.com/svn/trunk@13980 2bbb7eff-a529-9590-31e7-b0007b416f81
Use path rendering to render the text from outlines if supported by the
GPU. Implement this in GrStencilAndCoverTextContext by copying large
chunks of code from GrBitmapTextContext (drawText) and
GrDistanceFieldTextContext (drawPosText).
The drawing is implemented with "instanced" path drawing
functions.
Moves the creation of the "main" text context from SkGpuDevice to the
GrContext::createTextContext. This is done because the decision of which
text renderer is optimal can be made only with the internal
implementation-specific information of the context.
R=jvanverth@google.com, bsalomon@google.com
Author: kkinnunen@nvidia.com
Review URL: https://codereview.chromium.org/196133014
git-svn-id: http://skia.googlecode.com/svn/trunk@13962 2bbb7eff-a529-9590-31e7-b0007b416f81
This will be used in Blink to accommodate matrices that contain
rotation or shearing. This is a generalization of SkResizeImageFilter,
so I've replaced all uses of SkResizeImageFilter in Skia. (It might be
easier to review by diffing it with SkResizeImageFilter, too.)
R=reed@google.com
Review URL: https://codereview.chromium.org/211103006
git-svn-id: http://skia.googlecode.com/svn/trunk@13941 2bbb7eff-a529-9590-31e7-b0007b416f81
When gathering information about a picture (in the new SkDevice::EXPERIMENTAL_optimize entry point) it is necessary to be able to correlate the gathered information with the command in the SkPicture (so the information can later be combined with the similarly indexed information from the BBH). This entry point exposes that information to friend classes.
R=reed@google.com
Author: robertphillips@google.com
Review URL: https://codereview.chromium.org/206853003
git-svn-id: http://skia.googlecode.com/svn/trunk@13919 2bbb7eff-a529-9590-31e7-b0007b416f81
Tested in Mozilla source tree. I tried to use skia build system, but it's up to the task for cross compilation.
SkHRESULT.cpp - Use proper file name (that matters on case sensitive OSes)
SkAtomics_win.h - Don't use pragma intrinsic on GCC (this causes massive warnings)
SkOSFile_win.cpp - This one is tricky. GCC doesn't allow (void*) casts in template argument constants and INVALID_HANDLE_VALUE looks like this:
((HANDLE)(LONG_PTR)-1)
where HANDLE is typedefed to void*. Changed the code to use LONG_PTR as template argument and cast it when needed.
BUG=skia:
R=bungeman@google.com, reed@google.com
Author: cjacek@gmail.com
Review URL: https://codereview.chromium.org/198643004
git-svn-id: http://skia.googlecode.com/svn/trunk@13862 2bbb7eff-a529-9590-31e7-b0007b416f81
When accessBitmap() is called on a GPU device which has deferred clear,
and no drawing has been done, force the clear before returning the
bitmap.
This is covered by the top right test case in the imagefilterscropexpand
GM, where an offset image filter is drawn but the resulting bitmap is
cropped out, resulting in no drawing to the offscreen. It should be
empty, but without this fix it results in garbage (previous scratch
texture contents).
BUG=skia:
R=bsalomon@google.com
Review URL: https://codereview.chromium.org/204903002
git-svn-id: http://skia.googlecode.com/svn/trunk@13860 2bbb7eff-a529-9590-31e7-b0007b416f81
For the "pull forward" task I will be comparing the two cases:
analyze the whole skp and use the BBH information
analyze only the active portion of the skp
In the first case we need a way to get the BBH information out of the picture in order to extract the relevant portions of the whole-skp analysis. This adds caching of the active ops so that work isn't duplicated between when the optimization path queries for that information and when the usual draw path queries for it.
R=reed@google.com, bsalomon@google.com
Author: robertphillips@google.com
Review URL: https://codereview.chromium.org/195793010
git-svn-id: http://skia.googlecode.com/svn/trunk@13836 2bbb7eff-a529-9590-31e7-b0007b416f81
NOTE: this patch set is based on https://codereview.chromium.org/189913021/,
and needs that patch to land first.
Until now, crop rects in Skia have only been able to reduce
the size of the destination bounds, but not expand them.
SVG semantics require the latter as well. The heart of
the change is in applyCropRect(), which now assigns each
edge, instead of doing an intersection with the crop rect.
In order to support this (and still work well with tiled
drawing) we need to clip the resulting crop rect to the
clipping region of the filters. This uses the Context struct
previously landed from https://codereview.chromium.org/189913021/.
Many of the pixel loops are not yet ready to handle a
destination rect larger than the source rect. So we provide
a convenience version of applyCropRect() which creates an
offscreen and pads it out with transparent black. Once the
pixel loops and shaders have been fixed to support larger
destination bounds, they should be switched back to the
non-drawing version of applyCropRect().
BUG=skia:
R=bsalomon@google.com, reed@google.com
Committed: https://code.google.com/p/skia/source/detail?r=13805
Review URL: https://codereview.chromium.org/198003008
git-svn-id: http://skia.googlecode.com/svn/trunk@13809 2bbb7eff-a529-9590-31e7-b0007b416f81
NOTE: this patch set is based on https://codereview.chromium.org/189913021/,
and needs that patch to land first.
Until now, crop rects in Skia have only been able to reduce
the size of the destination bounds, but not expand them.
SVG semantics require the latter as well. The heart of
the change is in applyCropRect(), which now assigns each
edge, instead of doing an intersection with the crop rect.
In order to support this (and still work well with tiled
drawing) we need to clip the resulting crop rect to the
clipping region of the filters. This uses the Context struct
previously landed from https://codereview.chromium.org/189913021/.
Many of the pixel loops are not yet ready to handle a
destination rect larger than the source rect. So we provide
a convenience version of applyCropRect() which creates an
offscreen and pads it out with transparent black. Once the
pixel loops and shaders have been fixed to support larger
destination bounds, they should be switched back to the
non-drawing version of applyCropRect().
BUG=skia:
R=bsalomon@google.com, reed@google.com
Review URL: https://codereview.chromium.org/198003008
git-svn-id: http://skia.googlecode.com/svn/trunk@13805 2bbb7eff-a529-9590-31e7-b0007b416f81
Some upcoming work (support for expanding crop rects) requires
the clip bounds to be available during filter traversal. This change
replaces the SkMatrix parameter in the onFilterImage() traversals
with a Context parameter. It contains the CTM, as well as the clip
bounds.
BUG=skia:
R=reed@google.com
Review URL: https://codereview.chromium.org/189913021
git-svn-id: http://skia.googlecode.com/svn/trunk@13803 2bbb7eff-a529-9590-31e7-b0007b416f81
Fixed a few issues while attempting to use the new
serialization path for SkPicture inside a fuzzer:
- SkReadBuffer and SkValidatingReadBuffer both had a fReader
member instead of sharing the same member, which leads to
problems if a base class function is used
- In SkPicture, a header is now written as a single chunk of
data, so it also has to be read as a single chunk of data
- In the SkPicturePlayback destructor, a bad deserialization
would lead to a crash if we don't safely unref fOpData
- Also in SkPicturePlayback, if we only use a ReadBuffer for
the whole deserialization, additional tags must be added to
parseBufferTag()
- SkValidatingReadBuffer::readBitmap() was broken, but this
path wasn't usen't since the only use case for
SkValidatingReadBuffer is currently image filters and
bitmaps are unflattened as part of the deserialization of
SkBitmapSource
- SkPictureImageFilter was not deserializable. Added it to
SkGlobalInitialization*
- Added a test that exercises the SkPicture serialization /
deserialization code
BUG=skia:
R=senorblanco@google.com, senorblanco@chromium.org, reed@google.com, robertphillips@google.com
Author: sugoi@chromium.org
Review URL: https://codereview.chromium.org/195223003
git-svn-id: http://skia.googlecode.com/svn/trunk@13764 2bbb7eff-a529-9590-31e7-b0007b416f81
