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
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;client.skia.compile:Build-Mac10.8-Clang-Arm7-Debug-Android-Trybot
Committed: https://skia.googlesource.com/skia/+/a7627dc5cc2bf5d9a95d883d20c40d477ecadadf
Review URL: https://codereview.chromium.org/1317233005
The main thrust of this CL is to remove knowledge of the atlas' backing texture from the BatchPlot. It also reduces the API surface of the BatchPlot and shrinks the amount of fields in the BatchAtlas.
Review URL: https://codereview.chromium.org/1413403009
Reason for revert:
master-skia unhappy:
e6c439e806/logs/build_error.log
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
>
> CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;client.skia.compile:Build-Mac10.8-Clang-Arm7-Debug-Android-Trybot
>
> Committed: https://skia.googlesource.com/skia/+/a7627dc5cc2bf5d9a95d883d20c40d477ecadadfTBR=msarett@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:4117
Review URL: https://codereview.chromium.org/1409843005
Disable SkImageDecoder's code which relies on Android's customized
libpng and libjpeg. Build standard versions of libpng and libjpeg-turbo
everywhere. The SkImageDecoder code has been replaced with SkCodec, which
can decode subsets using standard library APIs
BUG=skia:
Review URL: https://codereview.chromium.org/1406153015
It has been demonstrated that higher quality settings
really do make a difference in the visual quality of
the output image.
https://code.google.com/p/chromium/issues/detail?id=385515https://code.google.com/p/skia/issues/detail?id=3770
We are planning to replace SkImageDecoder with SkCodec,
and SkCodec will use the higher quality settings. As
a first step, we are using SkCodec as the underlying
implementation for BitmapRegionDecoder. CTS tests require
that BitmapRegionDecoder be a close match to BitmapFactory
(which uses SkImageDecoder), so we must also update the
quality of SkImageDecoder to maintain CTS compatibility.
BUG=skia:
Review URL: https://codereview.chromium.org/1412803009
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
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;client.skia.compile:Build-Mac10.8-Clang-Arm7-Debug-Android-Trybot
Review URL: https://codereview.chromium.org/1317233005
We no longer need to worry about namespace
conflicts SkBitmapRegionDecoder in Android (which
we are replacing).
Additionally, the static Create() function does not
need to repeat the name BitmapRegionDecoder.
BUG=skia:
Review URL: https://codereview.chromium.org/1415243007
Quits requiring EXT_raster_multisample and
NV_sample_mask_override_coverage for mixed samples support. This will
allow platforms without those latter extensions (i.e. Chrome) to still
use mixed samples for path rendering. Also moves the mixed samples cap
out of shader caps, since it no longer denotes shader functionality.
BUG=skia:
Review URL: https://codereview.chromium.org/1410383011
This is the GPU equivalent of https://codereview.chromium.org/1415653003/.
It requires passing down the bounds of the crop rect (srcBounds), and
turning the blur 3-patch optimization in convolve_gaussian() into a 5-patch:
clear above and below srcBounds, blur with bounds checks inside left and
right rects, blur without bounds checks in middle rect.
Note: this change causes minor pixels diffs in the
imagefilterscropexpand GM: for odd crop positions relative to the
dstBounds, we are now correctly resampling at an even pixel boundary.
BUG=skia:4502, skia:4526
Review URL: https://codereview.chromium.org/1431593002
There seem about a zillion possible ways to slice this.
This adds and uses SkMallocPixelRef::ZeroedPRFactory.
I'm not married to it.
This appears to clear up a hot spot in the benchmark referenced in this bug:
BUG=516426
I'm not confident enough on Windows to declare that definitively yet.
It probably helps all other non-opaque layers too. Possibly significantly.
So I don't forget, I profiled this:
out/Release/performance_browser_tests.exe --gtest_filter=TabCapturePerformanceTest.Performance/2 --single-process
No diffs:
https://gold.skia.org/search2?issue=1430593007&unt=true&query=source_type%3Dgm&master=falseTBR=reed@google.com
Review URL: https://codereview.chromium.org/1430593007
Avoid hang in OncePtr test when using "dm --threads 1".
The test will hang the threads until sk_num_cores() threads have run
the code. This requires that sk_num_cores() threads to be run in
parallel, which the global thread pool will not do if the thread count
is smaller than sk_num_cores().
BUG=skia:
Review URL: https://codereview.chromium.org/1419593004
Make the struct passed to SkAndroidCodec::getAndroidPixels const. This
matches SkCodec, and makes sense, since it is not used as an output.
Brought up in crrev.com/1417583009
Review URL: https://codereview.chromium.org/1411693005
Remove double negative.
Also, change another error message to be slightly (meaningfully)
different so I can find the right one based on the log.
Review URL: https://codereview.chromium.org/1424083004
Reason for revert:
failing on msaa
Original issue's description:
> Revert[2] of "stop using drawSprite (at least w/ no filters) as it is going away"
>
> Fixed assert in test that was calling GrRecordReplaceDraw(). That function now uses
> drawBitmap, so updated the test to check for that (instead of drawSprite).
>
> This reverts commit 21b7663470.
>
> BUG=skia:
> TBR=
>
> Committed: https://skia.googlesource.com/skia/+/7b81994d95f10cda80ddb85af68a2651ff31782f
TBR=
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:
Review URL: https://codereview.chromium.org/1420053009
SkScalerContext_FreeType::generateCharToGlyph and generateGlyphToChar
do not take a mutex when calling into FreeType, but they need to.
The setupSize method also requires the mutex to be locked, which is
not a problem since it currently always is, but add a debug assert
to ensure that it continues to be.
BUG=chromium:542640
Review URL: https://codereview.chromium.org/1431683006
