Bug: skia:
Change-Id: I91b9816aae74726762c123d9f3454c5961382b7b
Reviewed-on: https://skia-review.googlesource.com/c/164680
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
The distinction between SkJumper and SkRasterPipeline used
to be important, but it's no longer. This CL moves everything
under src/jumper to the appropriate SkRasterPipeline file.
Change-Id: I1181fffafccb3dc4c4eb5f33b442c719ee370462
Reviewed-on: https://skia-review.googlesource.com/c/164627
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Same deal as before, just swap_rb where needed.
The change to SkWebpCodec could probably land independently if we want.
Change-Id: Idd53ab76232cb95eb2f41cd65c6903fa7c5b01d6
Reviewed-on: https://skia-review.googlesource.com/c/163440
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
We only need to decode the source file once, not multiple times.
Bug: skia:
Change-Id: I1067eac6b7acc482f97fb1835c1c219d21cc37e9
Reviewed-on: https://skia-review.googlesource.com/c/163300
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Dropping the "r" makes it consistent with both the class name and
Benchmark::getName.
Bug: skia:
Change-Id: I17bcf699d162c678daabf92b4e39b7230428b98b
Reviewed-on: https://skia-review.googlesource.com/c/163280
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This adds SkPMColor4f conversions to/from RGBA bytes (ie GrColor).
I had previously made some free functions that did the same thing.
I'm ambivalent about which option is nicer, but wanted to have one
method, so I converted everything to use the new versions.
Bug: skia:
Change-Id: I4194c44b5bd12228075fd1932a14cf31c8d6a3c1
Reviewed-on: https://skia-review.googlesource.com/c/162560
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:
Change-Id: I2cfc365e41942ad7cc3fbdf9b815df6e25b1f36f
Reviewed-on: https://skia-review.googlesource.com/c/160601
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Coping paints take significant time. The ApplyFontToPaint
idiom requires a copy and a dtor. This CL keeps the paint and
font in parallel through the code until a paint is actually
needed, then a special ctor is used to create it.
Also, inline a bunch of text blob calls that were showing up
in perf.
Change-Id: I7da746a287e4d3942e45e9536ef9acdc64f084d4
Reviewed-on: https://skia-review.googlesource.com/c/159222
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Herb Derby <herb@google.com>
It's been driving me nuts that I can't just write `SkMatrix44 m;`,
and I often don't care whether it's initialized or not. The default
identity constructor would be nice to use, but it's deprecated.
By tagging this constructor deprecated, we're only hurting ourselves;
our big clients disable warnings about deprecated routines and use it
freely.
A quick tally in Skia shows we mostly use the uninitialized constructor,
but sometimes the identity constructor, and there is a spread of all
three in Chromium. So I've left the two explicit calls available.
I switched a bunch of calls in Skia to use the less verbose constructor
where it was clear that it didn't matter if the matrix was initialized.
Literally zero of the kUninitialized constructor calls looked important
for performance, so the only place I've kept is its lone unit test.
A few places read clearer with an explicit "identity" to read.
Change-Id: I0573cb6201f5a36f3b43070fb111f7d9af92736f
Reviewed-on: https://skia-review.googlesource.com/c/159480
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:
Change-Id: Ie59aace6ba7ca3685d481fcb3af508629c56f0c3
Reviewed-on: https://skia-review.googlesource.com/157742
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
I keep seeing it show up on the profile(usually under memmove) of tight
benchmarks and it's kind of distracting. We don't even print it when
we pass --quiet, so that seems like a nice way to stifle it.
Change-Id: I3a67a7ca1758fd35e3b63cfeeddeac4ff1ffe38d
Reviewed-on: https://skia-review.googlesource.com/157520
Auto-Submit: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
- gammaencodedpremul GM was just demonstrating something that we
understand well (and have much better testing for).
- readpixels GM was filled with workarounds for things that are no
longer true (unpremul images, clamped F16).
- Other uses can be switched to SkConvertPixels trivially.
- Remove SkColorSpaceXformPriv and SkColorLookUpTable, all unused.
- Remove SkColorSpaceXform_skcms.cpp, no longer referenced by clients.
Bug: skia:
Change-Id: I7298bb53aa61b49ad1398ebc504d35c119fd5cf4
Reviewed-on: https://skia-review.googlesource.com/157153
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This is a reland of f3ebd312f2
PS 2: use faster half->float routine
PS 3: relax tolerance to 2^-10 for half, keeping others 2^-12
Original change's description:
> add getAlphaf() to pixmap/bitmap
>
> Convenient for just extracting alpha (and more efficient than getColor()) and
> works for super-normal formats w/o loss of precision.
>
> Somewhat inspired by examining multiple chrome call-sites for getColor(), where
> chrome only really cared about the alpha. This new method runs about twice
> as fast as getColor() for the simple cases (i.e. no colorspace xforms), and
> even faster in the complex cases (since retrieving alpha doesn't care about
> colorspaces).
>
> Bug: skia:
> Change-Id: I7cd5a2c7f78de781aaa69dd1aa0dba3c532fcb73
> Reviewed-on: https://skia-review.googlesource.com/155606
> Commit-Queue: Mike Reed <reed@google.com>
> Reviewed-by: Mike Klein <mtklein@google.com>
Cq-Include-Trybots: skia.primary:Test-Android-Clang-Nexus7-CPU-Tegra3-arm-Debug-All-Android,Test-Debian9-Clang-GCE-CPU-AVX2-x86_64-Release-All-SK_CPU_LIMIT_SSE2,Test-Debian9-GCC-GCE-CPU-AVX2-x86_64-Release-All,Test-Win2016-MSVC-GCE-CPU-AVX2-x86_64-Release-All,Test-Win8-Clang-Golo-CPU-AVX-x86_64-Debug-All
Bug: skia:
Change-Id: Ie94e5c89e185fde12cbd6c56ed4026c4dc5a1623
Reviewed-on: https://skia-review.googlesource.com/156242
Commit-Queue: Ravi Mistry <rmistry@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Nothing's using it except test tools.
I'd like to make that a bit clearer by getting it out of src.
Disabled the fuzzer.
Removed the bench so Android's building nanobench doesn't block this.
Bug: chromium:886713
Change-Id: I761f52c40171c27ff4b699409b32647e84684ec3
Reviewed-on: https://skia-review.googlesource.com/156240
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
This is a reland of 6d0e566e94
On second thought, it's probably better to correct the
types of the swizzle functions to express their required alignment.
This is a more involved CL, but I think leaves things better off.
Original change's description:
> have SkConvertPixels use SkColorSpaceXformSteps
>
> This ought to allow the fast paths in more cases, e.g. memcpy() when
> both src and dst are the same format. Today if we tag a dst color space
> at all, we'll think we need to fall back to the general case pipeline.
>
> Some refactoring too, but no big functional change beyond using steps.
>
> Change-Id: I8fa01025229e3b9418e7f43241a2f03628a97288
> Reviewed-on: https://skia-review.googlesource.com/155640
> Reviewed-by: Brian Osman <brianosman@google.com>
> Commit-Queue: Mike Klein <mtklein@google.com>
Change-Id: Ia17d93acfe88a36c4c36d29e3a0b243f91178b61
Reviewed-on: https://skia-review.googlesource.com/156241
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This reverts commit f3ebd312f2.
Reason for revert: unhappy bots
Original change's description:
> add getAlphaf() to pixmap/bitmap
>
> Convenient for just extracting alpha (and more efficient than getColor()) and
> works for super-normal formats w/o loss of precision.
>
> Somewhat inspired by examining multiple chrome call-sites for getColor(), where
> chrome only really cared about the alpha. This new method runs about twice
> as fast as getColor() for the simple cases (i.e. no colorspace xforms), and
> even faster in the complex cases (since retrieving alpha doesn't care about
> colorspaces).
>
> Bug: skia:
> Change-Id: I7cd5a2c7f78de781aaa69dd1aa0dba3c532fcb73
> Reviewed-on: https://skia-review.googlesource.com/155606
> Commit-Queue: Mike Reed <reed@google.com>
> Reviewed-by: Mike Klein <mtklein@google.com>
TBR=mtklein@google.com,reed@google.com
Change-Id: I82ce00c09d16bf3e9b04f1c1bccd8cc6aa706ab2
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/156000
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Convenient for just extracting alpha (and more efficient than getColor()) and
works for super-normal formats w/o loss of precision.
Somewhat inspired by examining multiple chrome call-sites for getColor(), where
chrome only really cared about the alpha. This new method runs about twice
as fast as getColor() for the simple cases (i.e. no colorspace xforms), and
even faster in the complex cases (since retrieving alpha doesn't care about
colorspaces).
Bug: skia:
Change-Id: I7cd5a2c7f78de781aaa69dd1aa0dba3c532fcb73
Reviewed-on: https://skia-review.googlesource.com/155606
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Bug: skia:
Change-Id: I62f02a122b431e6c9a866e80c441f3e18b1e69fb
Reviewed-on: https://skia-review.googlesource.com/155842
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
This reverts commit 90d2d9381e.
Updated to use float in the shader (many ES2 implementations
don't support integral types as attributes).
Bug: skia:
Change-Id: I0898f1730da60ff03f8165b2f1a3ee18a7b2fec8
Reviewed-on: https://skia-review.googlesource.com/155162
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
On my Z840, Windows-Clang-Debug, this cuts the total time to construct
(parse) the 72 SVG sources from 66 seconds to 40 seconds. That's still
awful, but all the time is now spent in expat, so further improvements
will require higher level changes.
Bug: skia:
Change-Id: I0dca67ee18652f6fb8647fe8706716d9a01f7cdf
Reviewed-on: https://skia-review.googlesource.com/155603
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This reverts commit d2ca0473f6.
Reason for revert: Various bots failing.
Original change's description:
> Add Short4 vertex attributes, and benchmark them as 4.12 colors
>
> This is likely our best widely portable option for encoding wide gamut
> colors that doesnt sacrifice too much performance. The benchmark uses
> them as we're likely to do: 4.12 fixed point, in the destination color
> space. We're using SINT vertex attributes for simplicity, so the encode
> and decode are simple multiply/divide by 4096.
>
> Bug: skia:
> Change-Id: I9b544f3e187b775d81f83dc9dd44611570ad33c2
> Reviewed-on: https://skia-review.googlesource.com/155001
> Reviewed-by: Mike Klein <mtklein@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
TBR=mtklein@google.com,bsalomon@google.com,brianosman@google.com
Change-Id: I921e29cf9a121c5752ec7d99a03470b5193afd4a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/154883
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This is likely our best widely portable option for encoding wide gamut
colors that doesnt sacrifice too much performance. The benchmark uses
them as we're likely to do: 4.12 fixed point, in the destination color
space. We're using SINT vertex attributes for simplicity, so the encode
and decode are simple multiply/divide by 4096.
Bug: skia:
Change-Id: I9b544f3e187b775d81f83dc9dd44611570ad33c2
Reviewed-on: https://skia-review.googlesource.com/155001
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Updates the existing vertex color bench with a half-float mode.
Vanilla GL2/ES2 don't have half-float attributes. The extension
(OES_vertex_half_float) uses a different enum value for the type
parameter. For now, just support GL3/ES3. Otherwise, we could
add a check in GrGLGpu::setupGeometry to re-map GL_HALF_FLOAT
to GL_HALF_FLOAT_OES, based on caps?
Bug: skia:
Change-Id: I1325087db4e615023cceffe1c5b051232769d0ac
Reviewed-on: https://skia-review.googlesource.com/154822
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
The CPU type is still specified using GrVertexAttribType.
The GPU type is specified directly using GrSLType.
kHalfX_GrVertexAttribType now really means half-float buffer
data, rather than float. (Caveat: The GL enum is only correct
with ES3/GL3 - ES2+extension needs a different value. Sigh.)
Bug: skia:
Change-Id: Ife101db68a5d4ea1ddc2f6c60fbec0c66d725c16
Reviewed-on: https://skia-review.googlesource.com/154628
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
New templated function SkPDFMakeArray that for example replaced this
code:
auto array = sk_make_sp<SkPDFArray>();
array->reserve(4);
array->appendInt(0);
array->appendInt(0);
array->appendInt(width);
array->appendInt(height);
with this code:
auto array = SkPDFMakeArray(0, 0, width, height);
Move some functions from SkPDFUtils to the only place they
are used and make them static: SkPDFUtils::AppendTransform,
SkPDFUtils::DrawFormXObject, SkPDFUtils::WriteString, and
SkPDFUtils::AppendCubic.
Also replaced SkPDFResourceDict::getResourceName with
SkPDFResourceDict::WriteResourceName and eliminated
SkPDFResourceDict::GetResourceTypePrefix.
Change-Id: I891339fa2d1e5819f22fb8d10d8d8ef75c9507e3
Reviewed-on: https://skia-review.googlesource.com/153884
Commit-Queue: Hal Canary <halcanary@google.com>
Reviewed-by: Leon Scroggins <scroggo@google.com>
This reverts commit 8b68110507.
Reason for revert: breaks bots
Original change's description:
> SkStream: remove some WStream functions from public api
>
> move functions to SkStringPriv.h
>
> also add SkStrAppendU32Hex() function, and re-write
> SkString::insertHex() to use SkStrAppendU32Hex.
>
> add unit tests.
>
> Change-Id: Ieda98fb4106db71565b607e593713a91a5ddd892
> Reviewed-on: https://skia-review.googlesource.com/151986
> Reviewed-by: Cary Clark <caryclark@google.com>
> Commit-Queue: Hal Canary <halcanary@google.com>
TBR=halcanary@google.com,bungeman@google.com,caryclark@google.com,reed@google.com
Change-Id: Idbac615092f46c18b38e08385dafba20930f0ff0
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/152121
Reviewed-by: Cary Clark <caryclark@skia.org>
Commit-Queue: Cary Clark <caryclark@skia.org>
move functions to SkStringPriv.h
also add SkStrAppendU32Hex() function, and re-write
SkString::insertHex() to use SkStrAppendU32Hex.
add unit tests.
Change-Id: Ieda98fb4106db71565b607e593713a91a5ddd892
Reviewed-on: https://skia-review.googlesource.com/151986
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Hal Canary <halcanary@google.com>
The code generated from std::move()ing them should be fine.
Bug: skia:8355
Change-Id: I63ef650b5fbcf9fb6356006190eae5e0977ae642
Reviewed-on: https://skia-review.googlesource.com/151982
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
All PDFs are identical.
Change-Id: If2af8519a6440a5b61f91bf78fc7ae241f728183
Reviewed-on: https://skia-review.googlesource.com/151220
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
Auto-Submit: Hal Canary <halcanary@google.com>
This reverts commit 5f7b5e3624.
Reason for revert: Codec CL has re-landed.
Original change's description:
> Revert "Stop conflating F16 with linear gamma"
>
> This reverts commit d1589c7213.
>
> Reason for revert: Depends on skcms CL that's been reverted.
>
> Original change's description:
> > Stop conflating F16 with linear gamma
> >
> > Note to self: I debugged this, realized that the codecs
> > need to handle A2B -> XYZ, then realized that I just need
> > to wait for https://skia-review.googlesource.com/c/skia/+/136062
> >
> > Bug: skia:
> > Change-Id: I594c22076feb3700b8a40c471a541fef5ff4e13e
> > Reviewed-on: https://skia-review.googlesource.com/137587
> > Commit-Queue: Brian Osman <brianosman@google.com>
> > Reviewed-by: Mike Klein <mtklein@google.com>
>
> TBR=mtklein@google.com,brianosman@google.com
>
> Change-Id: I6dca583697c8efd2563d30cb7ab9ef505b6903ae
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:
> Reviewed-on: https://skia-review.googlesource.com/148860
> Reviewed-by: Brian Osman <brianosman@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
TBR=mtklein@google.com,brianosman@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: skia:
Change-Id: Iee66531049843758e7ed4130b99d8df6a553d805
Reviewed-on: https://skia-review.googlesource.com/149700
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This reverts commit d1589c7213.
Reason for revert: Depends on skcms CL that's been reverted.
Original change's description:
> Stop conflating F16 with linear gamma
>
> Note to self: I debugged this, realized that the codecs
> need to handle A2B -> XYZ, then realized that I just need
> to wait for https://skia-review.googlesource.com/c/skia/+/136062
>
> Bug: skia:
> Change-Id: I594c22076feb3700b8a40c471a541fef5ff4e13e
> Reviewed-on: https://skia-review.googlesource.com/137587
> Commit-Queue: Brian Osman <brianosman@google.com>
> Reviewed-by: Mike Klein <mtklein@google.com>
TBR=mtklein@google.com,brianosman@google.com
Change-Id: I6dca583697c8efd2563d30cb7ab9ef505b6903ae
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/148860
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Note to self: I debugged this, realized that the codecs
need to handle A2B -> XYZ, then realized that I just need
to wait for https://skia-review.googlesource.com/c/skia/+/136062
Bug: skia:
Change-Id: I594c22076feb3700b8a40c471a541fef5ff4e13e
Reviewed-on: https://skia-review.googlesource.com/137587
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
We can now pass all cs/at params, so no need for UnpremulToUnpremul().
Also found a spot to use sk_srgb_singleton().
Does this make sense to apply to GrColorSpaceXformEffect too?
Change-Id: I1ee1f5919aa2ae841a98eeb5662cb80c244b1049
Reviewed-on: https://skia-review.googlesource.com/148665
Auto-Submit: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Various renames and other refactorings that will allow us to add new
stroking classes alongside the existing code for fills.
Bug: skia:
Change-Id: Ib477f9e1d87f9d4c1604719f9af0695a53614081
Reviewed-on: https://skia-review.googlesource.com/147503
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
All the existing calls are ported over, and new tests cover
unpremul output. (Opaque output is an impossible request.)
Change-Id: I744d640763cf74c368d3b3aba4a262c8fd9f7a01
Reviewed-on: https://skia-review.googlesource.com/147100
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: Mike Klein <mtklein@google.com>
Spotted the VisualBench hooks the other day, then found some more dead
code when I went to remove them.
Change-Id: Ia07f0556faf33cc69e3ec590f7a8f47f2c1f298a
Reviewed-on: https://skia-review.googlesource.com/146360
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
I hate SkColorSpaceXform,
and we think this is going to be faster, especially for single colors.
Add this mode of operation to our bench from yesterday.
No surprise, it's exactly as fast as pipeA.
This may DO the TODO?
Change-Id: I1b42b42d7647d85b62cd8397bb8d6f39c4829bcf
Reviewed-on: https://skia-review.googlesource.com/145823
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
w/ mtklein suggestions
Change-Id: Ifd7310ad71020c23514c663a0671400112b04e75
Reviewed-on: https://skia-review.googlesource.com/145824
Commit-Queue: Mike Klein <mtklein@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Auto-Submit: Herb Derby <herb@google.com>
PipeA is where we want to be, and probably by filling in something
like the TODO I left in SkColorSpaceXformSteps.h?
curr/maxrss loops min median mean max stddev samples config bench
13/13 MB 49 82.8ns 83.2ns 83.2ns 84.1ns 0% █▂▅▄▂▁▃▄▃▃ nonrendering ColorSpaceXformBench_pipeB
13/13 MB 1259 29.6ns 29.7ns 30.4ns 37.2ns 8% ▁▁▁▁▁▁▁▁▁█ nonrendering ColorSpaceXformBench_pipeA
13/13 MB 661 110ns 119ns 119ns 128ns 4% ▅▅▅▅▅▅▅▅█▁ nonrendering ColorSpaceXformBench_steps
13/13 MB 26 239ns 247ns 283ns 609ns 40% ▁▁█▁▁▁▁▁▁▁ nonrendering ColorSpaceXformBench_xform
Change-Id: I98af634ab0a7a1522a5bcfde9ce3fc934d82ecf8
Reviewed-on: https://skia-review.googlesource.com/145643
Commit-Queue: Mike Klein <mtklein@google.com>
Auto-Submit: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This reverts commit fdf05f4ff4.
Reason for revert: Android fixed after removing multitexture support from TextureOp.
Original change's description:
> Revert "Move GrGeometryProcessor's textures out of classes and into"
>
> This reverts commit af87483873.
>
> Revert "GrGeometryProcessor derives from GrNonAtomicRef not GrProgramElement."
>
> This reverts commit 607be37e3d.
>
> Revert "Store GrMeshDrawOps' meshes in GrOpFlushState's arena."
>
> This reverts commit b948572c78.
>
> Revert "Remove multitexturing support from GrTextureOp."
>
> This reverts commit 986f64c601.
>
> Revert "Make result of GrOp::combineIfPossible be an enum."
>
> This reverts commit 641ac7daa8.
>
> Bug: b/112244393
> Change-Id: I579491a3f2f2f2093f1e2a6141fa1e4cc7b760a4
> Reviewed-on: https://skia-review.googlesource.com/145646
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
TBR=bsalomon@google.com
Change-Id: I1d41c2ecf7862e31fb025a7a00bb07bae9d83a47
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: b/112244393
Reviewed-on: https://skia-review.googlesource.com/145780
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit af87483873.
Revert "GrGeometryProcessor derives from GrNonAtomicRef not GrProgramElement."
This reverts commit 607be37e3d.
Revert "Store GrMeshDrawOps' meshes in GrOpFlushState's arena."
This reverts commit b948572c78.
Revert "Remove multitexturing support from GrTextureOp."
This reverts commit 986f64c601.
Revert "Make result of GrOp::combineIfPossible be an enum."
This reverts commit 641ac7daa8.
Bug: b/112244393
Change-Id: I579491a3f2f2f2093f1e2a6141fa1e4cc7b760a4
Reviewed-on: https://skia-review.googlesource.com/145646
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This is to prepare for a third value that requests that ops be linked
together so that the first op may do the work for multiple linked ops
without actually merging the GrOp objects.
Change-Id: Ib6e012a89be5edd054aee69d8475bea612331852
Reviewed-on: https://skia-review.googlesource.com/145522
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: Brian Salomon <bsalomon@google.com>
This turned out to be an optimization mostly for draws with small device
space areas. Moreover, to be an optimization rather than deoptimization
requires complicated per-GPU tuning where even different devices within
the same architecture require different tuning and tuning is different
between GL and VK.
We've decided to go another direction where we *don't* coalesce draws
but rather make it possible to switch textures quickly from within an
op. This should be a GPU-independent optimization that is also
independent of device space area covered.
Replaces the multitexturing benchmarks with a pair of benchmarks.
composting_images* simulates a layered tile-based compositor.
image_cycle draws tiny images N times each such that they can be
batched. This is to catch a particular possible regression in a
planned change to GrTextureOp where it will use "dynamic state" to
batch across textures. We want to catch a slowdown that might result
from putting draws that use the same texture into the same op but no
longer batching the actual GL level draws and instead using the dynamic
state to "switch" textures between draws.
Change-Id: Ib1cc437525b0b0d56969c30dcb66bb1effb42dc5
Reviewed-on: https://skia-review.googlesource.com/145423
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Remove late draw consolidation in GrOpFlushState. Rarely did anything
and doesn't work with new allocation strategy. Ops can use GrMesh arrays
to acheive the same thing. (Each Op that cared to would have to implement
but it isn't applicable to most Ops).
Modify GrMeshDrawOp::Target::draw() to take array of meshes, with single
mesh as a special case.
Change-Id: I552677de47b9ffd2fcaf55af85f70f290e5aa9c7
Reviewed-on: https://skia-review.googlesource.com/145426
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>