Also, remove the never-used gpu_threading flag. In limited
testing, it makes things slower.
Change-Id: I40ae0c8e5b79992c90845035a06536ed9d6626b5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/240202
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
The client can do a test run of their application with
a persistent cache set to SkSL mode. They store the key
and data blobs that are produced.
Ship those blobs with the application. At startup, call
GrContext::precompileShader for each key/data pair. This
compiles the shaders, and stores the GL program ID, plus
a small amount of metadata in our runtime program cache.
Caveats:
* Currently only implemented for the GL backend. Other
backends will require more metadata to do any useful
amount of work. Metal may need a more drastic workflow
change, involving offline compilation of the shaders.
* Currently only implemented for cached SkSL (not GLSL
or program binaries). Supporting other formats again
requires more metadata, and the cached shaders become
increasingly specialized to GPU and driver versions.
* Reusing the cached SkSL on different hardware is not
supported. Many driver workarounds are implemented in
the SkSL -> GLSL transformation, but some are higher
level. Limiting device variance by artificially hiding
extensions may help, but there are no guarantees.
* The 'gltestprecompile' DM config exercises this code
similarly to 'gltestpersistentcache', ensuring that
results are visually identical when precompiling, and
that no cache misses occur after precompiling.
Change-Id: Id314c5d5f5a58fe503a0505a613bd4a540cc3589
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/239438
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This reverts commit a7b224d3a3.
Reason for revert: Failures on some ANGLE bots
Original change's description:
> DM: Add a per-sink MemoryCache to all GPU sinks
>
> This isolates the caches per config, so we don't try to reuse cached
> binaries across different GL versions (or different D3D backend versions
> on ANGLE bots, etc).
>
> Locally, this sped up ANGLE significantly Looking at a sample bot
> (NUC6i5SYK), it dropped DM from 36m to 31m.
>
> Change-Id: I95db18c6ea85041e6adbc972f0118233c82818fb
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/223985
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
TBR=bsalomon@google.com,brianosman@google.com
Change-Id: I8c9af65d89fddb28e5f1395a6c946312f27ba48a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/224186
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This isolates the caches per config, so we don't try to reuse cached
binaries across different GL versions (or different D3D backend versions
on ANGLE bots, etc).
Locally, this sped up ANGLE significantly Looking at a sample bot
(NUC6i5SYK), it dropped DM from 36m to 31m.
Change-Id: I95db18c6ea85041e6adbc972f0118233c82818fb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/223985
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
It's been dead code since it was inlined into Android last summer.
Change-Id: I252f6392d9436ef357f22a54bab8a33c9d1b3ea9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/213625
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Current strategy: everything from the top
Things to look at first are the manual changes:
- added tools/rewrite_includes.py
- removed -Idirectives from BUILD.gn
- various compile.sh simplifications
- tweak tools/embed_resources.py
- update gn/find_headers.py to write paths from the top
- update gn/gn_to_bp.py SkUserConfig.h layout
so that #include "include/config/SkUserConfig.h" always
gets the header we want.
No-Presubmit: true
Change-Id: I73a4b181654e0e38d229bc456c0d0854bae3363e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209706
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Hal Canary <halcanary@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
We need to store (up to) three GLSL strings in the cache entry,
along with the bookkeeping to reconstruct them. To make things
simpler, we now store the null terminators.
Change-Id: Ic4fe03cb5d774464372ceec8740da1bfe9069550
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/205823
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
- SkColorSpaceXformer
- makeColorSpace on SkShader, SkColorFilter,
SkImageFilter, SkDrawLooper, and SkLights
- DM support and some bot configs
Bug: skia:8773
Change-Id: I16ef8f487de6c35329b3b0474c1d66d7fa0a6220
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/202430
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Reed <reed@google.com>
The command line flag package is tool-only, not part of Skia per se,
and does not need an Sk prefix to avoid naming conflicts.
And git clang-format.
Change-Id: Ida8477779e51750ed0475590ed2454841b23d6ea
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/202307
Auto-Submit: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Replace all the old modes with two that should draw the same:
1) decode to the codec's "native" SkImageInfo, draw to dst;
2) decode to the dst's ct/at/cs, draw to dst.
Testing against --config srgb, p3, rec2020, narrow ought to give
us the kind of destination variety we were getting from the old
modes, and --configs 8888 shows off managed vs. unmanged color.
The only difference between the two is that 1) does any conversion at
draw time, 2) at decode time.
I don't quite understand why codec won't decode to some images to 565,
nor any images to 4444.
Bug: skia:8816
Change-Id: Ie57cb57312fdf21fd8dc36a0fa4d2c3b22e77acf
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/198181
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Clears between the two DDL runs.
Also, fix up proxy unique key checks in GrSurfaceProxy.
Change-Id: I492e791ebc57a42063f3b828c10d8bf5fee70b1b
Reviewed-on: https://skia-review.googlesource.com/c/178262
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
One less SkCanvas subclass to deal with...
Change-Id: I21e81648026be5d732e8d9a28baed55015492a04
Reviewed-on: https://skia-review.googlesource.com/c/161584
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This allows DM to use the same source in multiple threads.
Change-Id: Ia0d7c4c673a92ccddab263a67f01afaa2d68b606
Reviewed-on: https://skia-review.googlesource.com/144420
Auto-Submit: Florin Malita <fmalita@chromium.org>
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
Commit-Queue: Florin Malita <fmalita@chromium.org>
It is unused, is becoming a maintainence burden and source of bugs,
and takes up a lot of time on the *SAN bots.
Change-Id: If383eb6e4838ca23140f9e16d518b1bfc655fa12
Reviewed-on: https://skia-review.googlesource.com/143307
Auto-Submit: Mike Klein <mtklein@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Uses a new GPU sink that runs each test twice, once to populate the
cache and then again with a new GrContext but a warmed cache. It
verifies that the two generated images are the same.
Change-Id: Iaba195a69751f14ea946afe7174228a813b83a63
Reviewed-on: https://skia-review.googlesource.com/140567
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Most of this is (obviously) not necessary to do, but once
I started, I figured I'd just get it all. Tools (nanobench,
DM, skiaserve), all GMs, benches, and unit tests, plus support
code (command line parsing and config stuff).
This is almost entirely mechanical.
Bug: skia:
Change-Id: I209500f8df8c5bd43f8298ff26440d1c4d7425fb
Reviewed-on: https://skia-review.googlesource.com/131153
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Most of this CL is just repackaging the promise image and tile
code from ViaDDL for reuse by SKPBench.
Change-Id: Ie5003c36fe85cc5be9639552f9488b8e92dcdbbf
Reviewed-on: https://skia-review.googlesource.com/129805
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Make Skottie truly optional (own flag), and disable in framework builds
(to unblock landing the RapidJson refactoring).
Change-Id: I4611f915e43fe11c1f6754ab4a9f63e45af2f8d3
Reviewed-on: https://skia-review.googlesource.com/125872
Commit-Queue: Florin Malita <fmalita@chromium.org>
Reviewed-by: Mike Klein <mtklein@google.com>
This reverts commit ea51393ba9.
Reason for revert: try again after many fixes of the threaded backend
Original change's description:
> Revert "Exercise the threaded backend in test bots"
>
> This reverts commit a39991ebd7.
>
> Reason for revert: Random bots timeout
>
> Original change's description:
> > Exercise the threaded backend in test bots
> >
> > We can't draw everything correctly now, but it's at least not crashing.
> >
> > The draw_to_canvas is modified by adding flush because now the raster
> > canvas also needs flush like GPU canvases because of the threaded
> > backend. (Previously, AndroidCodecSrc vetoed GPU (non-raster) sink
> > and that's why it was not crashing.)
> >
> >
> > Bug: skia:
> > Change-Id: I222d706c33e15326e1b97d6bef6c46b8c35cebfd
> > Reviewed-on: https://skia-review.googlesource.com/118886
> > Reviewed-by: Kevin Lubick <kjlubick@google.com>
> > Commit-Queue: Yuqian Li <liyuqian@google.com>
>
> TBR=kjlubick@google.com,stephana@google.com,liyuqian@google.com
>
> Change-Id: Ib1089d937b6e25cc39d90ad4de35b3a4c406f6b1
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:
> Reviewed-on: https://skia-review.googlesource.com/119100
> Reviewed-by: Yuqian Li <liyuqian@google.com>
> Commit-Queue: Yuqian Li <liyuqian@google.com>
TBR=kjlubick@google.com,stephana@google.com,liyuqian@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: skia:7838
Change-Id: I1405212204bf694b2f31dfc38154759e1d3e363f
Reviewed-on: https://skia-review.googlesource.com/121680
Commit-Queue: Yuqian Li <liyuqian@google.com>
Reviewed-by: Yuqian Li <liyuqian@google.com>
Change-Id: I6bb270d9df1c76b9d2e384abde603cdda91c9bb6
Reviewed-on: https://skia-review.googlesource.com/116550
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This will let us also render the GMs via DDLs.
TBR=mtklein@google.com
Change-Id: If7c2460d964822a6decc33cf5e8e685e67923127
Reviewed-on: https://skia-review.googlesource.com/116463
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This reverts commit bd2e2a0caf.
Reason for revert:
Build-Debian9-Clang-x86_64-Debug-MSAN
In file included from ../../../dm/DMSrcSink.cpp:8:
../../../dm/DMSrcSink.h:548:15: error: private field 'fNumDivisions' is not used [-Werror,-Wunused-private-field]
const int fNumDivisions;
Original change's description:
> Switch DDL rendering to be a Via in DM
>
> This will let us also render the GMs via DDLs.
>
> Change-Id: I866a5af66d737473f4760dbc9d45053460c42e6e
> Reviewed-on: https://skia-review.googlesource.com/116021
> Commit-Queue: Robert Phillips <robertphillips@google.com>
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Reviewed-by: Mike Klein <mtklein@google.com>
TBR=egdaniel@google.com,mtklein@google.com,robertphillips@google.com
Change-Id: Ice07c25ab29c57b2efb193c03e5504a2ae1a0743
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/116560
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This will let us also render the GMs via DDLs.
Change-Id: I866a5af66d737473f4760dbc9d45053460c42e6e
Reviewed-on: https://skia-review.googlesource.com/116021
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This also adds GrGpu::create/deleteTestingOnlyBackendRenderTarget. Implemented in GL only for now.
Change-Id: I9e5fdc953c4a249959af89e08332f520cefe9d90
Reviewed-on: https://skia-review.googlesource.com/113305
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Currently with dm, it's possible to convert an .mskp to a multi-page PDF as
follows:
out/Release/dm --src mskp --mskps /tmp/filename.mskp -w /tmp \
--config pdf --verbose
The SVG equivalent partially works, although only outputs the first page:
out/Release/dm --src mskp --mskps /tmp/filename.mskp -w /tmp \
--config svg --verbose
This CL adds support for passing extended options to SVG. Specifically, the
'page' option , which now determines which page of the source mskp gets
converted to the SVG output. The new syntax is as follows:
out/Release/dm --src mskp --mskps /tmp/filename.mskp -w /tmp \
--config svg[page=2] --verbose
The `[key=value]` syntax is the same extended options syntax currently used by
dm with --config gpu, e.g. `gpu[api=gl,color=8888]`.
BUG=skia:7601
Change-Id: I3523d79b1cdbbba9e80fd46501331877091bdead
Reviewed-on: https://skia-review.googlesource.com/105404
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
The bug is spotted by the new sink. I'd expect more bugs to be revealed
and fixed using the new sink.
Bug: skia:
Change-Id: I6b0c9267079fbd6149004f0ecd55ddb179702588
Reviewed-on: https://skia-review.googlesource.com/109720
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
Extend composition layers to support referencing external .json
animations ("$"<PATH> syntax).
This is a custom extension (not supported in BM/Lottie).
Also make skottie::Animation ref-counted, to facilitate sharing.
TBR=
Change-Id: I062d031e5868d759f3930dea9b261f9b3ec81684
Reviewed-on: https://skia-review.googlesource.com/109806
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Florin Malita <fmalita@chromium.org>
This reverts commit 3a2cc2c2ec.
Fix code with samplecnt=0 that slipped in between trybots/CQ and landing of previous version
Change-Id: Iab19f2e8d1e9901601c8c76244d7a88c5d707fab
Reviewed-on: https://skia-review.googlesource.com/103181
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit 5bb82cbecd.
Reason for revert: <INSERT REASONING HERE>
Original change's description:
> Revert "Revert "Revert "Revert "Redefine the meaning of sample counts in GPU backend.""""
>
> This reverts commit 18c52a7b52.
>
> Also relands "More sample count cleanup:" and "Add new GrContext queries for imagability, surfacability, and max sample count of color types"
>
>
> Bug: skia:
> Change-Id: I4028105a3a1f16ce3944e134619eb6245af6b947
> Reviewed-on: https://skia-review.googlesource.com/102940
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
TBR=egdaniel@google.com,bsalomon@google.com
Change-Id: Idee23be2f1719f0bdc9305043e95a2d589bee8d1
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/103220
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit 18c52a7b52.
Also relands "More sample count cleanup:" and "Add new GrContext queries for imagability, surfacability, and max sample count of color types"
Bug: skia:
Change-Id: I4028105a3a1f16ce3944e134619eb6245af6b947
Reviewed-on: https://skia-review.googlesource.com/102940
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit d0d7270fcc.
Revert "More sample count cleanup:"
This reverts commit d653cac70e.
Revert "Add new GrContext queries for imagability, surfacability, and max sample count of color types"
This reverts commit 85ae7159c9.
Need to understand NVPR perf changes before relanding
Change-Id: I0db075fb42438ef2a1f9885df184dce52892ac4b
Reviewed-on: https://skia-review.googlesource.com/102780
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Fixes gpu config default samples to be 1 and updates config parsing test accordingly.
This reverts commit c1ce2f7966.
Bug: skia:
Change-Id: I456973b1f52ced85a2011ea10fc49449bfc5846f
Reviewed-on: https://skia-review.googlesource.com/102147
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit 48825b11ad.
Reason for revert: nanobench
Original change's description:
> Redefine the meaning of sample counts in GPU backend.
>
> Old: 0 -> nonMSAA
> 1+ -> MSAA
>
> New:
> 0 -> error/unsupported
> 1 -> nonMSAA
> 2+ -> MSAA
>
> We still allow 0 to mean nonMSAA in three sets of public APIs for backwards compatibility:
>
> 1) SkSurface factories
> 2) GrBackendRenderTarget constructors
> 3) GrCaps::getSampleCnt()'s requestedCount parameter
>
> However, we immediately clamp to 1 and treat 0 as invalid/non-renderable internally.
>
> This also changes the behavior when using a large sample count. We now fail in that case rather than using the largest sample available sample count. GrCaps::getSampleCount() will return 0 in this case.
>
>
> Bug: skia:
> Change-Id: Ida22c6b22c1365e563c9046b611e88bf5eb3ff33
> Reviewed-on: https://skia-review.googlesource.com/101560
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Brian Salomon <bsalomon@google.com>
TBR=egdaniel@google.com,bsalomon@google.com
Change-Id: Ic257619a8a5ee9ac15419ecf10259e42daed7f82
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/102662
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Old: 0 -> nonMSAA
1+ -> MSAA
New:
0 -> error/unsupported
1 -> nonMSAA
2+ -> MSAA
We still allow 0 to mean nonMSAA in three sets of public APIs for backwards compatibility:
1) SkSurface factories
2) GrBackendRenderTarget constructors
3) GrCaps::getSampleCnt()'s requestedCount parameter
However, we immediately clamp to 1 and treat 0 as invalid/non-renderable internally.
This also changes the behavior when using a large sample count. We now fail in that case rather than using the largest sample available sample count. GrCaps::getSampleCount() will return 0 in this case.
Bug: skia:
Change-Id: Ida22c6b22c1365e563c9046b611e88bf5eb3ff33
Reviewed-on: https://skia-review.googlesource.com/101560
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
ViaSingletonPictures simulates a load pattern that Blink no longer produces.
I don't even remember what Via2ndPic and ViaTwice are meant to test,
but I imagine they're quite obsolete.
There was a lingering reference to ViaMojo, which otherwise doesn't exist.
Bug: skia:7544
Change-Id: I033fd344314054ad66c363e5dbd0a373be188069
Reviewed-on: https://skia-review.googlesource.com/88780
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
To our knowledge it's not used.
Its defer- mode in DM isn't even run on the bots.
Change-Id: Ifebfa2a77bfed8370eb421d379697f04fa2c8608
Reviewed-on: https://skia-review.googlesource.com/76420
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Chrome would like to perform cpu-side preprocessing for gpu draws in parallel.
They do not want to go through a picture (since they have their own display list format).
The general idea is that we add a new SkDeferredDisplayListRecorder class to
perform all of Ganesh's cpu-side preprocessing ahead of time and in parallel.
The SkDDLRecorder operates like SkPictureRecorder. The user can get an SkCanvas
from the SkDDLRecorder and feed it draw operations. Once finished, the user
calls 'detach' to get an SkDeferredDisplayList. All the work up to and
including the 'detach' call can be done in parallel and will not touch
the GPU. To actually get pixels the client must call SkSurface::draw(SkDDL)
on an SkSurface that is "compatible" with the surface characterization
initially given to the SkDDLMaker.
The surface characterization contains the minimum amount of information Ganesh needs
to know about the ultimate destination in order to perform its cpu-side work
(i.e., caps, width, height, config).
Change-Id: I75faa483ab5a6b779c8de56ea56b9d90b990f43a
Reviewed-on: https://skia-review.googlesource.com/30140
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Bug: skia:
Change-Id: Icd905ea7e60a05bc3903eb85d111dcf73ce2c4dd
Reviewed-on: https://skia-review.googlesource.com/40690
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
