Additional shrinking of GrRenderTargetOpList since there will soon be more of them.
Change-Id: Ib3e68fdf3462838baf7a5a2644e418be066cf79c
Reviewed-on: https://skia-review.googlesource.com/14363
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Bug: skia:6516
Change-Id: Iec940c585654d0c45aa8de7743c270bd54f6af7a
Reviewed-on: https://skia-review.googlesource.com/14370
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Oddly, the MSVC toolchain already says "assemble ...".
Change-Id: Iabf06afa1b7c05b5699a26813f1469346be439a5
Reviewed-on: https://skia-review.googlesource.com/14367
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
If/when we double check these settings we can either re-add these parameters or change the code manually.
The driving goal here is to shrink GrRenderTargetOpList.
Change-Id: I7819de18f97e8282e8754e3438130f797ebb3c37
Reviewed-on: https://skia-review.googlesource.com/14381
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
The parametric_{r,g,b} stages are just as good now;
under the hood it's all going through approx_powf.
Change-Id: If7f3ae1e24fcee2ddb201c1d66ce1dd64820c89a
Reviewed-on: https://skia-review.googlesource.com/14320
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
All users have now been updated and this define is no longer used.
Change-Id: Ib701b6e31f7c0958eed3d97e033b334e09fa9bb0
Reviewed-on: https://skia-review.googlesource.com/14365
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
This also now requires an explicit vulkan proc address getter to be used with GrVkBackendContext::Create.
Change-Id: I768ec487398dab68e6e10409419ac7a00ec1822a
Reviewed-on: https://skia-review.googlesource.com/14322
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This wasn't seen locally b.c. it is an assert and I only ran release locally and the CQ also only runs release.
I have added linux_trusty_blink_dbg as a try job.
TBR=bsalomon@google.com
Bug: 715392
Change-Id: I010626cb97e886d2fbfd767f948bc640f0534338
Reviewed-on: https://skia-review.googlesource.com/14361
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
SkNoncopyable declares (but does not define) its copy constructor and
copy assignment operator. These are also private so the error for misuse
happens at compile-time instead of link-time.
However, this seems to be from before C++11.
Because other constructors were declared, the compiler does not generate
a move constructor or a move assignment operator. The result of this is
perfectly legal non-copying scenarios are also accidentally blocked. An
example of this is returning the non-copyable type. The object being
returned is a candidate for a move, since it is about to be destroyed.
And in C++17 copy elision is actually guaranteed.
Change-Id: Ia31be9091c644f31a45dd18216330a68be3cf456
Reviewed-on: https://skia-review.googlesource.com/14294
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
BUG=chromium:710131
Change-Id: I4568bc24cc8fabb6f9df3b7645b01a98e11791b8
Reviewed-on: https://skia-review.googlesource.com/14062
Reviewed-by: Doug Anderson <dianders@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
I missed this while removing SkRasterPipeline_opts.h.
It's just a declaration... this won't change any generated code.
Change-Id: I66f6e9fe5341e9ff6a91981da9275c944a63fee9
Reviewed-on: https://skia-review.googlesource.com/14325
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Adjusted unit test to verify this behavior.
Bug: skia:6547 chromium:713632 chromium:713702
Change-Id: I6240937b2faf6ccb6adfc9477dc85ae961cdbbb7
Reviewed-on: https://skia-review.googlesource.com/14279
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Bug: skia:
Change-Id: I6b46f6e448b52a4473c5c16342ebe002b6818c8a
Reviewed-on: https://skia-review.googlesource.com/14301
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Stephan Altmueller <stephana@google.com>
We now will always use the system vulkan.h files whenever we are building
with vulkan. With non vulkan builds we use our checked in header to so
that we can get the needed symbols for compiling.
Bug: skia:
Change-Id: I352a3e007b33c575cefcfd6752db0b3b12b86a16
Reviewed-on: https://skia-review.googlesource.com/14270
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Before, we would stomp on the original paint in the recursive call.
This fixes 4 gbr-8888 gms. Currently, this only affects loopers
and SkPaintImageFilter.
Bug: skia:6516
Change-Id: Ic47d637a912370c0a1ae8ef3282ad7d15d9902e3
Reviewed-on: https://skia-review.googlesource.com/14182
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
All variants of (on)?[rR]efEncoded(Data)? no longer need a GrContext
parameter.
Bug: skia:5485 skia:4971
Change-Id: If4f5e785718d5522eb3df8588318ccb8a02a5749
Reviewed-on: https://skia-review.googlesource.com/14269
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
CTFontCopyVariationAxes returns nullptr for fonts which began life with
CGFontCreateWithDataProvider on macOS 10.10 and earlier and iOS 9 and
earlier. The issue was fixed in maxOS 10.11 and iOS 10. Improve the
existing comments to make this clear.
Change-Id: I0de1734e266d30aab7d38d0dd752f66259f677d9
Reviewed-on: https://skia-review.googlesource.com/14278
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Just noticed this TODO is easy now. This doesn't seem to change any
GMs, and definitely isn't used by Chrome or Android... this is just for
completeness.
It's also definitely correct to unfurl pictures.
Change-Id: Ida6e43788d1de7a9452dad47aea8a5fbe41a1f46
Reviewed-on: https://skia-review.googlesource.com/14276
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Matt Sarett <msarett@google.com>
Adds more jobs to iOS testing and switches to
RPi based hosts.
Bug: skia:
Change-Id: I0124c6554173fa3b22b077a198612bb30691d1fb
Reviewed-on: https://skia-review.googlesource.com/11521
Commit-Queue: Stephan Altmueller <stephana@google.com>
Reviewed-by: Eric Boren <borenet@google.com>
Bug: skia:6516
Change-Id: Id09336def938f6e4da028807912494cdae5f3734
Reviewed-on: https://skia-review.googlesource.com/14190
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Matt Sarett <msarett@google.com>
* Use unique_ptr, sk_sp, SkAutoFree, using.
* Rely on thread-safe static global initializion.
Change-Id: I7c14e0e57622163b1b81b97a218b816fe6d02926
Reviewed-on: https://skia-review.googlesource.com/13818
Commit-Queue: Hal Canary <halcanary@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Better imitate the original Android bug. Create a stream with
multiple images in it, and verify that it successfully decodes after
decoding once.
This exposes a bug in SkPngCodec, which did not work for interlaced
images.
Test more formats that also happen to succeed: ICO, BMP, and WBMP
This explicitly does *not* attempt to fix sampled or subset
decodes, which already stopped early when decoding as an
optimization.
Change-Id: Ib0b8918f14ba3fb0fa31e9c71c8100dcbeeb465f
Reviewed-on: https://skia-review.googlesource.com/14104
Reviewed-by: Matt Sarett <msarett@google.com>
Also remove the feature of GrVkGpu that creates the instance/device if the client doesn't provide one.
Change-Id: Ie617313b6c684ed355333a475b80d0aae7e3a026
Reviewed-on: https://skia-review.googlesource.com/14261
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
- lockAsBitmap supported reading back textures to a bitmap, but this was
only used by one GM. Removed all of that code, and merged the two bitmap
functions together.
- To make the GM cleaner, don't use SkImageCacherator directly - construct
actual images.
Bug: skia:
Change-Id: Iad17184a02a72f89ccc851fa158764d6c871befa
Reviewed-on: https://skia-review.googlesource.com/14192
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
When determining whether or not to use symmetric rendering,
only look at the gasp table if it's version 1.
Change-Id: Ib64ec1853545eb2c467a9e280d9e00f51e8b0aa8
Reviewed-on: https://skia-review.googlesource.com/13863
Commit-Queue: Ben Wagner <bungeman@google.com>
Reviewed-by: Dominik Röttsches <drott@google.com>
Simplifies the process of nesting environments. Usage:
with api.env({'myvar': 'myval'}):
# do stuff
Same as api.step.context but specialized for just environment and merges
PATH variable intelligently.
Bug: skia:6473
Change-Id: I5769c69cbbbcdab0c6298cee6c5e1fe9caf89c78
Reviewed-on: https://skia-review.googlesource.com/14189
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Commit-Queue: Eric Boren <borenet@google.com>
This reverts commit df2bf21364.
Reason for revert: Maybe AndroidOne timing out
Original change's description:
> Split up opLists (take 2)
>
> Reland of: https://skia-review.googlesource.com/c/11581/ (Split up opLists)
>
> https://skia-review.googlesource.com/c/13860/ (Make InstancedRendering more opList-splitting friendly) has landed so this should be good for another attempt.
>
> Change-Id: Icc9998196587510328e0a9ca1b2ce42013a86c6c
> Reviewed-on: https://skia-review.googlesource.com/13802
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Robert Phillips <robertphillips@google.com>
>
TBR=bsalomon@google.com,robertphillips@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I744f2a3145b294e5911862bb39d57ca33a1b9a5a
Reviewed-on: https://skia-review.googlesource.com/14184
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
If process_data is unable to read (and therefore process) as many bytes
as it expects, process the bytes read before returning false.
Fixes differences in Gold.
Add a test that verifies that it is okay to call png_process_data with
0 bytes. (We could special case 0, but libpng already checks for 0.)
Change-Id: Id500b9305ee3bb6a1a7e8fc70d4e723cb4742b55
Reviewed-on: https://skia-review.googlesource.com/14144
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>
Bug: skia:
Change-Id: Ibc6698b4db8aabeb14224d2c3ec8412f7ecfecb9
Reviewed-on: https://skia-review.googlesource.com/14146
Reviewed-by: Brian Osman <brianosman@google.com>
This is probably it until GrTexture.h can be hidden.
TBR=bsalomon@google.com
Change-Id: Ie743c937f72aa0dc9f5e84def572f655f0abe041
Reviewed-on: https://skia-review.googlesource.com/14143
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This CL splits the InstancedRendering class into an allocator and a per-opList op tracker. This was done because we need to allocate the InstancedRendering ops before we know the final destination opList. The InstancedRendering ops are now still all allocated from the same pool but the tracking and execution is done per opList.
Change-Id: Ieddabb6f85f8f187c5e7373f7f6cb155d69a9685
Reviewed-on: https://skia-review.googlesource.com/13860
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Chris Dalton <csmartdalton@google.com>
As long as everything is laid out the same way they were originally, I don't
think there's any reason we can't just use %rip-relative addressing on x86-64.
Basically, we just need to keep all the sections together in order.
Somewhat subtly we cannot just use -D to disassemble all sections. -D will
double-disassemble[1] some bytes, which throws off our %rip-relative addressing
of constants. You can see this in PS1. So we whitelist sections instead.
[1], from man objdump:
This option also has a subtle effect on the disassembly of instructions in code
sections. When option -d is in effect objdump will assume that any symbols
present in a code section occur on the boundary between instructions and it will
refuse to disassemble across such a boundary. When option -D is in effect however
this assumption is supressed. This means that it is possible for the output of -d
and -D to differ if, for example, data is stored in code sections.
Change-Id: Idbcfe08e67113b3f7d75749931c640ff90aa0bf4
Reviewed-on: https://skia-review.googlesource.com/14029
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
(Originally uploaded as 13900.)
Previously, SkPngCodec assumed that the stream only contained one
image, which ended at the end of the stream. It read the stream in
arbitrarily-sized chunks, and then passed that data to libpng for
processing.
If a stream contains more than one image, this may result in reading
beyond the end of the image, making future reads read the wrong data.
Now, SkPngCodec starts by reading 8 bytes at a time. After the
signature, 8 bytes is enough to know which chunk is next and how many
bytes are in the chunk.
When decoding the size, we stop when we reach IDAT, and when decoding
the image, we stop when we reach IEND.
This manual parsing is necessary to support APNG, which is planned in
the future. It also allows us to remove the SK_GOOGLE3_PNG_HACK, which
was a workaround for reading more than necessary at the beginning of
the image.
Add a test that simulates the issue, by decoding a special stream that
reports an error if the codec attempts to read beyond the end.
Temporarily disable the partial decoding tests for png. A larger change
will be necessary to get those working again, and no clients are
currently relying on incrementally decoding PNGs (i.e. decode part of
an image, then decode further with more data).
Include a workaround for older versions of libpng (e.g. 1.2 in
Google3). In older versions, if the row callback is null when the
IDAT header is processed, reading the image will fail. When we see the
IDAT, we save the length and process a recreated IDAT header later,
after the row callback has been set.
Bug: skia:5368
Bug:b/34073812
Test: Existing tests, plus a new test in dm.
Change-Id: I293a4ddc013b82669a8b735062228b26d0bce933
Reviewed-on: https://skia-review.googlesource.com/13984
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>
These are slow enough that they sometimes trigger GPU timeout.
Bug: skia:6534
Change-Id: Ie16a533e5ae132599d6514ae15a22d4fce4cb0fd
Reviewed-on: https://skia-review.googlesource.com/14064
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Fixes 4 gms in gbr-8888 config.
Bug: skia:6516
Change-Id: I9da839eb0211910989be35db0e13c0e1bbfa185d
Reviewed-on: https://skia-review.googlesource.com/13964
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Matt Sarett <msarett@google.com>
Also adds an SkColorSpace to the BitmapShader
cache key (that should have been there already).
This fixes 3 gms in gbr-8888 config.
Bug: skia:
Change-Id: If47dfcdbd99b4caaeee4eddbe5ff1cd042ea9d02
Reviewed-on: https://skia-review.googlesource.com/13976
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Matt Sarett <msarett@google.com>
This (only?) affects ANGLE. Other ES platforms typically go through
EGLGLTestContext, which manually instantiates an EGLFenceSync.
In general, though, ES3 requires this API, so this is safe. Should give us
more accurate (and much less spammy) output from ES3 ANGLE performance
testing.
Bug: skia:
Change-Id: I10a608d21092aaffa4ab76e4b3d2f6e9c5cf09bb
Reviewed-on: https://skia-review.googlesource.com/14063
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
