For approximate-fit deferred proxies asserts were firing when the instantiated size was larger than the pre-computed exact-fit size.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4972
Change-Id: I879e16b86ab4d9ef9834163c24ccd6507fe4b94a
Reviewed-on: https://skia-review.googlesource.com/4972
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
When VkPipeline creation fails the PipelineStateBuilder cleans up
afterwards. It was deleting the sampler DescriptorSetLayout when that
is owned by the ResourceProvider.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4964
Change-Id: I876c570349f96f58e0256dc91bc1135c59a2058d
Reviewed-on: https://skia-review.googlesource.com/4964
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Let's see if this can convince MSAN that we're safe.
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot,Perf-Ubuntu-Clang-GCE-CPU-AVX2-x86_64-Debug-MSAN-Trybot
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4971
Change-Id: Ie90124068bc412bee94a4f8e87fae32e4d66e709
Reviewed-on: https://skia-review.googlesource.com/4971
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
The Google3 Android ARM build is still using GCC, which isn't so loosey-goosey as Clang about bit-casting back and forth between int32x4_t and uint32x4_t.
BUG=skia:
Change-Id: I0d54f6859b8e03be4936c51bbaa1967db4261bd4
Reviewed-on: https://skia-review.googlesource.com/4974
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
In the past, ANGLE always used GL_HALF_FLOAT. I filed a bug about that,
and they switched to HALF_FLOAT_OES. However, ES3 has direct support for
HALF_FLOAT, so we expect that as the format for read pixels.
The better solution would be to remember and return the value we get
back from IMPLEMENTATION_COLOR_READ_TYPE, but there are subtleties that
create additional bugs when we do that. In particular:
If trying to read Alpha8 from RGBA, the current sequence is:
1. readPixelsSupported asks for the format to use. getExternalFormat
switches out the format (assuming we're using GL_RED). This creates
a format/type pair that fails the ES3.2 rules (16.1.2), so we ask
the driver for the secondary format, which matches the one that we
expect (from our config table).
2. Then we *actually* do the readPixels, and here we just call
getReadPixelsFormat, which does the remapping, but skips the follow
up checks in readPixelsSupported. So we end up passing a format/type
pair that our code in readPixelsSupported thinks is illegal (and
which is neither the pair that our config table stores, *nor* the
pair returned by the implementation).
A straightforward solution of the original problem is to notice that
we had to ask the implementation (because the values are filled in),
so we should trust those and not the ones in the config table. But then
we miss out on the chance to overwrite the values for the alpha only
from RGBA case. Sigh.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4963
Change-Id: Icb0c151c9f07ffc178ed3a597fa8ab0ed3c6cb94
Reviewed-on: https://skia-review.googlesource.com/4963
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
First of many CLs, I'm sure.
This handles 8888 or sRGB sources with an affine matrix, clamp/clamp tiling, and nearest-neighbor sampling only.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4906
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Change-Id: I99f7508852b3d44b6f52f7a0bee29a793af35c48
Reviewed-on: https://skia-review.googlesource.com/4906
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This is a follow up to https://skia-review.googlesource.com/c/4929/ (Guard against instantiate & accessRenderTarget failures).
Rather than guard this call to accessRenderTarget I would prefer to remove it.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4961
BUG=665681,665500,665621
Change-Id: I2c9ec245491d9059de892b2e6a7d4a4de4accdfd
Reviewed-on: https://skia-review.googlesource.com/4961
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
It seems that SkFixedMul_lowprec doesn't have too much performance gain (maybe
1%). But its precision loss is siginificant. No serious issues have been found
in convex cases, but things get much trickier for concave shapes. I'm removing
it now to improve the stability and reliability of our algorithm, which may
potentially benefit our Chrome landing. (Even if we do not remove
SkFixedMul_lowprec now, we eventually will remove it when we land concave AAA
code.)
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4926
Change-Id: Iae5739a3780bb77ce6237888eee51a502fea5cf2
Reviewed-on: https://skia-review.googlesource.com/4926
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
Chrome's fuzzer have reminded me that, since we are deferring allocation, instantiate and accessRenderTarget can now fail further down the call stack.
This should probably be cherry picked back to M56.
BUG=665681,665500,665621
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4929
Change-Id: I44d81ff29586dfe75ddda30b5ed8ca76354542d6
Reviewed-on: https://skia-review.googlesource.com/4929
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Milestone that we typically update, but note that change from 55->56 was missed, will cherry pick to branch.
This change gets current file caught up to show working milestone after branch.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4927
Change-Id: If10e0db2d4acc908f5da7c4e9d30497deef6a2ff
Reviewed-on: https://skia-review.googlesource.com/4927
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Heather Miller <hcm@google.com>
Commit-Queue: Heather Miller <hcm@google.com>
Renames vars and methods that used the work "access" to refer to this type.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4931
Change-Id: Ibcf488fbd445c5119fc13d190544cd98981bdbee
Reviewed-on: https://skia-review.googlesource.com/4931
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
skia.h includes GPU headers. No real point in getting fancy here... if there are no unused parameters in all headers, there are no unused parameters in the subset of non-GPU headers.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4916
Change-Id: Iaaea405633aabcc82e40cd4b1a2286aadd5cea17
Reviewed-on: https://skia-review.googlesource.com/4916
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Introduce an SkImageGenerator API to support the implementation of
externally-managed image decode and scale caches.
BUG=skia:5806
R=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4720
Change-Id: Ibfe37af5471f78f28f88f9d5e80938882be1a344
Reviewed-on: https://skia-review.googlesource.com/4720
Commit-Queue: Florin Malita <fmalita@chromium.org>
Reviewed-by: Mike Reed <reed@google.com>
This also fixes the name of the tasks because skpbench is always GPU.
There was a config for monitoring the gpu clock instead of the cpu clock,
but this was removed because the gpu clock isn't very useful at the moment.
cpu clock is basically wall-time, i.e. an accurate measurement of the
entire test.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4873
Change-Id: If1c7dd3141e24f79555ee7b3756f09618df54cb4
Reviewed-on: https://skia-review.googlesource.com/4873
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Just a simple renaming to reinforce the Ganesh notion that a buffer
access is not related to a texture.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4879
Change-Id: I3f12b6eccff744e246618be0a7abf41c4a1228a1
Reviewed-on: https://skia-review.googlesource.com/4879
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
We already generate skia.h to include all public headers for Fiddle.
This just includes it with -Wunused-parameter turned on as an error.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4901
Change-Id: Ia55a901c09a3c9c9d6d35a43259431dba3532ed9
Reviewed-on: https://skia-review.googlesource.com/4901
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Yet more debugging tools.
TBR=herb@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4908
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Change-Id: I92e2e6b17abfc32c8d3554e71dbf95abadbb3824
Reviewed-on: https://skia-review.googlesource.com/4908
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
The virtual flush function doesn't seem to affect the performance much.
Maybe there's a 1% drop in performance in nanobench against
fill_big_triangle and fill_big_circle, but that's too small a change for
nanobench to reliabily diffrentiate.
The smooth jump (ignore fractional y if edges don't change their
directions significantly) no longer needs to be guarded against SkAAClip
because our recent CL (https://skia-review.googlesource.com/c/4636/)
make left/rightBound much tighter.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4899
Change-Id: If323013b810cc1ff5f6dbb868a8981354ee6f9b5
Reviewed-on: https://skia-review.googlesource.com/4899
Reviewed-by: Mike Reed <reed@google.com>
Reviewed-by: Stan Iliev <stani@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
Commit-Queue: Stan Iliev <stani@google.com>
Previously, the clipRect is either equal to nullptr or clipRgn's bound
(after necessary supersampling shift). Hence we drop one of them to make
the signature simpler.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4874
Change-Id: Ied8d5313809d6cf90374365b01f2b8d52f2236e2
Reviewed-on: https://skia-review.googlesource.com/4874
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
BUG=661244
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4738
Change-Id: I9f14ca830f9de92000e751a4a99ff1eb9b01db33
Reviewed-on: https://skia-review.googlesource.com/4866
Reviewed-by: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
This is step one:
- make SkXfermode useless to public clients
- everything they should need is in SkBlendMode.h
Step two:
- remove SkXfermode.h entirely (since skia core will already be using SkXfermodePriv.h)
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4534
Change-Id: If2cea9f71df92430ed6644edb98dd306c5572cbc
Reviewed-on: https://skia-review.googlesource.com/4534
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
Think you can take over from here, hook these into SkColorSpaceXform_A2B, and remove the need for fn_1_{r,g,b}?
BUG=skia:664864
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4888
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Change-Id: If573fa2861f32f201f4db28598559290b1eef812
Reviewed-on: https://skia-review.googlesource.com/4888
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Matt Sarett <msarett@google.com>
Fixes the sense of the "stencilEnabled" argument sent to
flushHWAAState.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=4872
Change-Id: If7097f2819607864c95b0ffd04dc1d302fa66fdc
Reviewed-on: https://skia-review.googlesource.com/4872
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
