(and from the sources, too)
Change-Id: I9d8ff51c91aad4b770b1f183c04734d31252b851
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313148
Commit-Queue: Julia Lavrova <jlavrova@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
Change-Id: Ia97c4d653331414a05e73945d3a9c6a953ffc97b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314497
Commit-Queue: Herb Derby <herb@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Auto-Submit: Herb Derby <herb@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Tiny diffs seen, but not sure how to avoid those and keep the code
path simple -- so just plan/hope to rebaseline as needed.
Change-Id: Id8ff7e85a6e70785592f76118a32def2d61599ec
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314076
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>
When sampler objects exist but we're not using them we bind 0.
Speculatively disable use on ANGLE to see effect on performance.
Change-Id: I5c76bd812760a5ac67a4327f2d02f4f831155029
Bug: skia:10644
Change-Id: I7ef23704b2ca11d92806f196589ec561a3ac9435
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314462
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
AtlasLocator maintains the page index, and atlas location in fUVs.
This reduces the number of instructions needed to produce the
glyphs vertices. Moving the page index to the high bits allows many
simplifications to the code because subtraction can be used to
calculate width and height, and small offsets can be added with
no re-encoding.
Change-Id: I592543f772274dd18e00c0aec2520aadb1bf4158
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314317
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Change-Id: Ief9aa9aa035424573533072af2c0cd8cee5bf9b4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314460
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This pattern is a bit more efficient when multiple removals occur; it
avoids moving elements up more than once.
(https://en.wikipedia.org/wiki/Erase%E2%80%93remove_idiom)
Change-Id: I612d3c52145c889e30f7203f3c7298e461e855a5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314459
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
No public user needs these.
Change-Id: Ieeb519a0778b01697c9bbb3b91ce8423eaa94e57
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314316
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This will allow the Compiler to access the Inliner (in addition to the
IRGenerator, as before).
Change-Id: I1aeeaf8e3d3fb5d15533f7bf5c635a4798115d1f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314357
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Ideally, we wouldn't need a counter at all here, and we could just rely
on the symbol table to check for name collisions. Unfortunately, this
fails in practice on a couple of tests, because we don't construct
programs in a strict top-to-bottom order--particularly when inlining.
However, by checking the symbol table before using a name, we can at
least solve cases where the inliner reuses a name that was taken in a
previous pass, which was why salt was originally added. This makes the
generated code much easier to read.
Change-Id: Ib69611c8df457fbd03b31e52158113ad4a8735d5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314277
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
It isn't actually fewer LOC, but the level of complexity is lower;
rather than having to reason about recursion, it allows the reader to
assume that ProgramVisitor does the right thing, and focus on the
interesting parts of the algorithm.
Change-Id: I3f110a6029481afac1c9634035eb275c3b2d6455
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314321
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Metal requires any variable in the vertex shader representing point size
to be assigned the [[point_size]] semantic. Otherwise, point size is
undefined.
Bug: skia:10658
Change-Id: Iac428175327a27ba319e9dab9a0c17d4b1731cce
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314318
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
If only a single statement is necessary, it is returned as-is.
If multiple statements are needed, they are wrapped in a non-scoped
Block statement. If no statements at all are needed (!), null will be
returned.
Change-Id: I6ba373f73d339b8c2e7b8d39dfb28f13655e3d03
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313911
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The functional changes needed to make this migration possible have
already been submitted in prior CLs, so although this is large, it is
almost entirely a mechanical migration of code from one file to another.
Change-Id: I54380b52e38ebcec40ffbca7cb254f9655cb1865
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313909
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This CL does several things:
1. SkDevice subclasses that implement drawSpecial() now assert that the
provided SkPaint does not have an SkImageFilter.
2. The SkGpuDevice implementation of drawSpecial() is simplified to go
through the draw_texture_producer code path. This removes duplicate logic
for paint and texture handling, and also allows simple drawSpecials to
actually use the texture op.
3. SkCanvas' drawInternalDevice and onDrawImage now define the
skif::Mappings that define the coord transforms for filter eval and
final draw, before calling the new drawFilteredImage on SkDevice.
- This is implemented in SkBaseDevice, but subclasses can extend it.
- I had originally put it as a static function in SkCanvas, but that
meant I had to forward declare more types in SkCanvas.h than I liked.
- A lot of the skif::Context properties are dependent on the dst device
anyways, so having a drawFilteredImage on device seems okay, now that
filter eval is separated from how to draw a special image.
- Once drawSpecial can take a matrix transform, and once image filters
know how to handle non-(0,0) source origins, this will simplify further.
4. Swapped the behavior of SkDevice::getRelativeTransform. I originally
wrote it so that it would be the matrix mapping from arg to caller.
After finally using it here (its original purpose), I realized it reads
better to be the matrix from caller to arg.
- Previously this was never called, so compatibility isn't a problem.
Bug: skia:9558
Change-Id: If855903110947a11ae89b1e50cf5848a3c5dbecd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/308768
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This doesn't change any behavior, just moves out repeated code to a
common place. (It does add a vector::reserve that wasn't present before,
but that's a pretty minor detail.)
Change-Id: I7b1fe8746af5192e2cddcc80418351395f3c702c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313838
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Use checkRange() to prevent an out-of-range enum.
Bug: oss-fuzz:25301
Change-Id: Icd20ebb8a60925d24ca39b9be0e85fafb361539a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314119
Auto-Submit: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>
This reverts commit 952f088d41.
Reason for revert: Removing temporary perf experiment hack
Original change's description:
> Experiment to see if ANGLE ES2 vs ES3 perf is related to GLSL ver
>
> Force ES2 level shading language on ANGLE.
>
> Will be reverted. Only affects Skia tool builds.
>
> Bug: skia:10644
> Change-Id: Ic52d6afae7f784b173725a9353d5f43f2da3ca72
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313703
> Commit-Queue: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Greg Daniel <egdaniel@google.com>
TBR=egdaniel@google.com,bsalomon@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: skia:10644
Change-Id: I5ba23c7479d6f7162a3b75e947787943866fb52f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/314176
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This moves us closer to the goal of allowing inlineCall() to inline
functions later in the compilation process, rather than only during
initial IR generation. (Redesigning how extra statements are injected is
another hurdle to overcome.)
Change-Id: Idee34579dcf4d4414080ef5885732f6f3406005a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313841
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Also, move first-direction into SkPathRef.h so it can be referenced
by name in SkPath (instead of using uint8_t)
No functional change expected.
Change-Id: Ica4a8357a8156fd9a516118f23599a965b0fdd47
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313980
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>
For some strange reason, bit 15 seems to create problems on iphone6
for gles. This works fine for mtl. Shift down to bits 13 and 14.
Change-Id: Id8e39d46bf23decaf1bd1a6058dd3fc999fc31cb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313907
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This flag was added recently since I noticed it was missing and should
be checked. However, Dawn fails this check in chrome since it doesn't
report this flag correctly to DDLs. Going back to the previous status
quo for now.
Bug: skia:10672
Change-Id: Ib825fe5a69bff7af0d9893b95cd4df80289be7b2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313905
Commit-Queue: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Auto-Submit: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Add API to atlas locator to manipulate the PlotLocator and rect.
The next step is to store the page index in the rect using the
update methods to maintain the invariants between the PlotLocator
and the rect.
Change-Id: I7e27d20ffce7f08f71fef35d02bbd59eb4235dae
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313903
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Store the 2 page index bits in bits 14 and 15 of U. This allows
several simplifications.
Change-Id: Ib3ce98f0f5a430849acc0e6f6f3ecfabdb3f038d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313896
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Subsequent CLs can work on combining convexity and direction
Bug: skia:10670
Change-Id: Ia44769ea88ffd99a56d4c6729a80a2044e790ec2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313837
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Mike Reed <reed@google.com>
This was only used by a single visitor (MergeSampleUsageVisitor) and
just for its context. It is simpler for MergeSampleUsageVisitor to just
hold onto a Context& directly.
Change-Id: Ia38e3fa1b67498ac4c65a6a8450ca87aea1e0963
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313876
Commit-Queue: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This sample attempts to re-create an image's alpha channel by drawing it
one pixel at a time and timing how long each pixel takes to draw.
The "abc" text should appear twice normally, and the third and fourth
versions are reconstructed from timing, one by timing 1:1 pixel draws,
the other by timing 1x1:1024x1024 upscale into an offscreen. It's not
meant to be an exact reconstruction, but you can easily see the shapes,
particularly at -O0, -O1, and -Os. Auto-vectorization from -O2/-O3 do
a good amount to cover up the problem.
The legacy CPU backend is the main place to look. I haven't been able
to reconstruct any images using SkRasterPipelineBlitter or SkVMBlitter,
and while on the GPU I do see non-random patterns in the timing, it
appears to be the same single pattern across devices, OSes, GPUs, GPU
APIs and content... I assume it's something like our resource caching
policy.
This can't really be a GM, given how it draws non-deterministically.
Bug: chromium:1088224
Change-Id: I2ec79c8dd407ecb104fd9bf0c8039cb6dd1fe436
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313466
Commit-Queue: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>
This doesn't change any behavior, just moves out repeated code to a
common place, and adds a handful of comments.
The code which creates a block of inlined statements has also been
reorganized a bit; the unique_ptr<Block> of inlined statements is now
created earlier, instead of creating a vector of statements and then
wrapping it in a Block in two separate places. Again, this shouldn't
cause any externally-visible behavioral changes; it's just doing the
same steps in a different order.
Change-Id: I89e87d93cf505ee30a49917390afab6a4f226aeb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313839
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
SkBlitRow_opts.h is including this and masking the issue.
Change-Id: I0a5f873a2f9c486d40f86e79c8119081495ae1ae
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313810
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This has always been a potential source of a bug. If the same texture is
used twice in a shader with different swizzles we would overwrite the
swizzle for the first use by that of the second use since there is
only one fixed function swizzle per texture. It's not part of the
sampler state.
We set the swizzle when it is a feature, but always to RGBA.
Also, highly speculative that this may improve ANGLE D3D11 ES3
performance compared to ES2.
Bug: skia:10644
Change-Id: I8877afc3043c5ddaafd26ea9f9bd372303328c71
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313682
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Using Chromium's clang with skia_use_metal=true produces
../src/gpu/mtl/GrMtlCommandBuffer.h:25:5: warning: '~GrMtlCommandBuffer' overrides a destructor but is not marked 'override' [-Wsuggest-destructor-override]
~GrMtlCommandBuffer();
Change-Id: I177a6645a37cf9853b9284798eb48997428e1854
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313808
Commit-Queue: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Auto-Submit: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Change-Id: I9b908bc1cdd28f3bd6819096420b1d119eecd62e
Bug: skia:10669
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313836
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: Joe Gregorio <jcgregorio@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
In a follow-up, I will explore more clearly tying together convexity
and direction internally -- perhaps a unified enum?
[ unknown, convex_cw, convex_ccw, concave ]
Change-Id: I9fc2a2205f40050f4c24c5bec7fc25c8b6d2461c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/313680
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>
