This drastically reduces the number of functions which we allow to be
inlined. If this change does not hurt our performance, it will allow us
to trivially remove hundreds of LOC. All current data leads us to
believe that it may affect the Mali 400 but is highly unlikely to change
results on any other device in the tree.
More info: http://go/optimization-in-sksl-inliner
Change-Id: Ia6b706742ce5407453e0e697b6c1f9201084c0e8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384858
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This test fails on SPIR-V when the inliner is turned off:
error: SPIR-V validation error: Pointer operand 250[%250] must be a memory object declaration
%252 = OpFunctionCall %void %out_half %250
Change-Id: Ibfa9cef371af2eea766a4218ec8a581289ee100e
Bug: skia:11748
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384999
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Prevents us from accepting code that can't be correctly transformed to
GLSL, like:
uniform float x;
float y = x;
(Previously, writing code like that in a runtime effect would
effectively produce the exact same code all the way through to GLSL, and
the driver would fail to compile it).
Bug: skia:11336
Change-Id: Iaa797587c4a4a7289ed59ce2736cf0bf0fc5bca3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384698
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
A side effect of this change is that I've tried to pass GrVkAttachments
around GrVkGpu instead of GrVkTextures where I could to start the
transition within the backend code.
Bug: skia:10727
Change-Id: Ibc9553cdbd7f6ae845c56aad3f25f58e4c478e46
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379577
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Bug: skia:11716
Change-Id: Ic09071544b5b5216b01fbc9b478b6269dd96202f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382280
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This only affects fragmentProcessors (children) - and the backend SkSL
we're emitting should not contain those. We've just been silently
ignoring those declarations when converting to GLSL, MSL, etc.
Change-Id: I241f2f4fe4614b49ebccc9c2976fd408e94656d0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384316
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This reverts commit a04692f69e.
Reason for revert: Angry Vulkan bots.
Original change's description:
> Fixed a number of spots where we should have been using RelaxedPrecision
>
> Our SPIR-V output was missing many RelaxedPrecision decorations, which
> was presumably impacting performance.
>
> Change-Id: Iee32d4a42f37af167fe0e45f3db94c2142129695
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384178
> Reviewed-by: Brian Osman <brianosman@google.com>
> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
TBR=egdaniel@google.com,brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: If4fe945cb363c9b61b5a4abfde649a437689d2eb
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384217
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
I ran into an issue in an upcoming CL which generated a particularly
ugly switch statement:
switch (x) {
default:
discard;}
So I cleaned this up, and while resolving this issue, managed to improve
a bunch of existing codegen as well. The formatting change has been
split out to a separate CL since it impacts so many golden outputs.
Change-Id: I7a6be29903c47560dcc7f6acd3ef15fd0c5c3c50
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384179
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Our SPIR-V output was missing many RelaxedPrecision decorations, which
was presumably impacting performance.
Change-Id: Iee32d4a42f37af167fe0e45f3db94c2142129695
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384178
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
All that's left is a clip atlas renderer.
Bug: chromium:1158093
Change-Id: I8b509904a752a202ff1321e5302c41a3f57a5edb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/383741
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Bug: skia:11738
Change-Id: I1dd5e99830f70d72c292379a45c4e39a55588858
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/383706
Commit-Queue: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
We no longer derive a performance benefit from this pass in practice,
and it is a very expensive compilation step. It is also prone to fuzz-
related errors.
Doc: http://go/optimization-in-sksl
Change-Id: Ief08ffac659a8fe7fe92c92b9a5da14c9f713bc2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381261
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Previously even if we release/abanoned a resource we would still return
GrBackendSurfaces from these queries even if the actual backend api
objects are no longer valid.
This hopefully fixes the attached chrome bug, but can't know for sure.
Bug: chromium:1186623
Change-Id: Ib1c03699c2c7d81f6d305428dfbf39d647d28373
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382918
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This declutters the atlas generation code so there are fewer variables.
Next we will delete all the lower level rendering code and render the
atlas with stencilPath/stencilRect instead.
Bug: chromium:1158093
Change-Id: I36cff285d0f7de6f8ece4b027e62ae84aa01adc8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380656
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Abandoning would cause us to unref all the command buffer usages, but
then in notifyARefIsZero we would delete the abandoned resources even
though we still had a real ref. Then when that ref went away we would
crash.
Change-Id: I05dc2ba9a67c35c36a36704f4b81d6eef4e860e6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382916
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
This is useful because it allows the variables to be declared as `const`
when they are trivial values like `half4(1)`. This enables the constant
folder to simplify or eliminate them. In most cases, this is only a
small benefit, as you'd expect a competent GPU driver to do the same.
However, Mali-400 can benefit significantly from optimizing away the
multiplication against a constant half4(1) coverage in Porter-Duff.
Mali-400 performance is back to normal: http://screen/3cDxdaGkYE8oBcS
Change-Id: I21fd23f91f747079cd05b082f7b3444aeabafb93
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382476
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
As you might expect, a function tagged with `noinline` will never be
considered as a candidate for inlining.
Change-Id: Ia098f8974e6de251d78bb2a76cd71db8a86bc19c
Bug: skia:11362
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382337
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Currently, only one of three uses (local variables) does this correctly.
Bug: skia:11716
Change-Id: Iad11e8e5998fcc7caee4d438e0558c5d4e2b1821
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382277
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
With skia-review.googlesource.com/c/skia/+/363782, both cpu and gpu
backends would gracefully fail readPixel requests where dst row bytes
wasn't a multiple of dst bpp. It also updated the cpu backend's
writePixels behavior to gracefully reject writePixels requests where
the src row bytes wasn't a multiple of src bpp.
GPU writePixels would not detect this and later trigger an assert
in debug builds in GrConvertPixels (caught by the linked fuzzer bug).
This adds tests to mirror the read pixels bad-row-bytes tests and
updates GrSurfaceContext::writePixels to check src row bytes vs. bpp.
I confirmed it fixes the fuzzer crash.
Bug: chromium:1185266
Change-Id: I7cd8406c65a9ba35a55d695b2f65410a1edd2a19
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/382276
Auto-Submit: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Change-Id: Ic2d1240ab785101365b0fd934562505fb5a3e599
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381816
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
It turned out that everywhere we were using or testing DSL code either
directly or indirectly imported big chunks of the SkSL library. These
imports turned out to be necessary; code written using just DSL.h would
fail with various template instantiation errors.
Change-Id: Iae72d15b0d6ef14614ac1a4ff08c36bc1876cd4d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381638
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
The crash is because method GrGpuBuffer::onRelease() is called on wrong
thread. SkMessageBus::Post(const Message& m) takes a const ref of the
message, so there is a little possibility the GPU thread received and
handled the message before SkMessageBus::Post() is returned. In that
case, the caller of SkMessageBus::Post() (AsyncReadResult) still holds
the last ref of the GrGpuBuffer, and then GrGpuBuffer() will just be
released on the wrong thread.
Bug: chromium:1185489
Change-Id: I28665dbb1db7925d59ec574e9e26385e845ff4df
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380696
Commit-Queue: Peng Huang <penghuang@chromium.org>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Auto-Submit: Peng Huang <penghuang@chromium.org>
These were accidentally omitted from the supported operator list.
Change-Id: Idecd17adb8b3f5043e36328c65ca12be33e990f4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381637
Auto-Submit: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Expressions like `value == value` or `color.a != color.a` can be
replaced by `true` or `false` on sight. The GLSL spec makes it clear
that checking for NaN is optional:
4.7.1 Range and Precision
"... NaNs are not required to be generated. Support for signaling NaNs
is not required and exceptions are never raised. Operations and built-in
functions that operate on a NaN are not required to return a NaN as the
result."
Change-Id: I2ad9f2dc505b638ea2904bef41b7a79a2b329551
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381262
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This CL will be used to test for potential performance regressions (or
improvements) that we might cause by disabling this optimization pass.
It will be reverted in ~1 day.
Change-Id: I26b7687c341eb6d81231406381c39869cfccf6d6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/381259
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
There's really only one failure at this point, and it's a failed
instantiation. The original intention of the allocation failure
system was to only drop the ops that referred to the uninstantiated
proxies, but somewhere along the way that was lost and
we started dropping arbitrarily large chunks of ops. Lets just
bail and not crash instead.
Bug: skia:10877
Change-Id: I675358e8a1fbd2d75ea29b72ccfc50c7df90343e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/371337
Commit-Queue: Adlai Holler <adlai@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Change-Id: I7a7874e58bf53978afce8a41b26092406b6490ed
Bug: skia:11342
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380360
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit 9ef3f2e3da.
Reason for revert: relanding with fix
Original change's description:
> Revert "Have GrVkRenderTarget only use GrVkAttachments and not derive from GrVkImage."
>
> This reverts commit 3dc6c190da.
>
> Reason for revert: hitting assert about RT having input attachment on mali bots
>
> Original change's description:
> > Have GrVkRenderTarget only use GrVkAttachments and not derive from GrVkImage.
> >
> > This change moves the color and resolve attachments used in a
> > GrVkRenderTarget to be a GrVkAttachment. These along with the msaa
> > attachment now mean that GrVkRenderTarget no longer needs to derive from
> > a GrVkImage.
> >
> > There are a couple ugly things in this CL since GrVkTexture still is a
> > GrVkImage since we can't share attachments between GrVkRT and GrVkTex.
> > But when that gets updated in the follow on CL things will look much nicer.
> >
> > Bug: skia:10727
> > Change-Id: I2f12674d7517c6d6dea389e2d1fb7296028bcc85
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379576
> > Reviewed-by: Brian Salomon <bsalomon@google.com>
> > Commit-Queue: Greg Daniel <egdaniel@google.com>
>
> TBR=egdaniel@google.com,jvanverth@google.com,bsalomon@google.com
>
> Change-Id: Ic46f3947ed9f7b2ca26e8418d643e7f89b6108d2
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:10727
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380459
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Greg Daniel <egdaniel@google.com>
# Not skipping CQ checks because this is a reland.
Bug: skia:10727
Change-Id: I7a995ee9ad35bdac34cfcfd6b0d963c3e0bb90b8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380460
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Expressions like `x * 1`, `x *= 1`, `x + 0`, `x * 0`, or `0 / x` don't
actually do anything, and can be simplified to just `x` or `0`. (The
zero case must also check that `x` doesn't have side effects, because
`0 * myFunction()` still needs to call `myFunction`.)
`0 - x` is also detected and rewritten as `-x`.
`0 / 0` is left as-is.
This logic works for scalars and vectors; matrices are left as-is.
A similar optimization also occurs in the constant-propagation pass, so
we see almost no diffs in the tests. If control-flow analysis is turned
off, we do see some improvements. (I didn't reuse the existing code at
all, since it was designed around rewriting the CFG tree, but the
concept was identical.)
Change-Id: Ia99cd81f1d4cd3dafaa43ccac6a2261e3257a185
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380356
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This reverts commit 3dc6c190da.
Reason for revert: hitting assert about RT having input attachment on mali bots
Original change's description:
> Have GrVkRenderTarget only use GrVkAttachments and not derive from GrVkImage.
>
> This change moves the color and resolve attachments used in a
> GrVkRenderTarget to be a GrVkAttachment. These along with the msaa
> attachment now mean that GrVkRenderTarget no longer needs to derive from
> a GrVkImage.
>
> There are a couple ugly things in this CL since GrVkTexture still is a
> GrVkImage since we can't share attachments between GrVkRT and GrVkTex.
> But when that gets updated in the follow on CL things will look much nicer.
>
> Bug: skia:10727
> Change-Id: I2f12674d7517c6d6dea389e2d1fb7296028bcc85
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379576
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Greg Daniel <egdaniel@google.com>
TBR=egdaniel@google.com,jvanverth@google.com,bsalomon@google.com
Change-Id: Ic46f3947ed9f7b2ca26e8418d643e7f89b6108d2
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:10727
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380459
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Declare
This solves several issues caused by the lack of ordering guarantees in
C++; it was possible for the SkSL backend to look for the value of a
variable before its Declare() call gets processed. Moving the initial
value out of Declare should fix this whole class of problems.
Change-Id: I428fe230f1c312a0128c1f00c2a36cb95f4590a6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/380358
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This enables the ternary to be optimized away in code like:
const bool SHINY = true;
color = SHINY ? add_shine(x) : x; // to --> `color = add_shine(x);`
Without constant propagation.
Also, I added a unit test for ternary expression simplification; I
wasn't able to find an existing one.
When the optimization flag is disabled, this CL actually removes the
optimization of `true ? x : y` --> `x` entirely; previously, this
substitution would be made regardless of optimization settings.
Change-Id: I93a8b9d4027902d35f8a19cfd6417170b209d056
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379297
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This change moves the color and resolve attachments used in a
GrVkRenderTarget to be a GrVkAttachment. These along with the msaa
attachment now mean that GrVkRenderTarget no longer needs to derive from
a GrVkImage.
There are a couple ugly things in this CL since GrVkTexture still is a
GrVkImage since we can't share attachments between GrVkRT and GrVkTex.
But when that gets updated in the follow on CL things will look much nicer.
Bug: skia:10727
Change-Id: I2f12674d7517c6d6dea389e2d1fb7296028bcc85
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379576
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Swizzle optimizations now occur at IR generation time. These
optimizations are redundant with the control-flow optimization phase so
they are mostly not visible in our test output, but they do affect DSL
test results. Interestingly, they do improve our test output slightly
as well, for various reasons (e.g. we do not fully optimize lvalues in
the control-flow pass).
Change-Id: I6ebe6d71a5c22d9823b5fa500e43078915cbfb45
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/372257
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Makes SkImage_Gpu backed by two proxies, an original and a copy. The
image uses the original until a new render task is bound to it at which
point further uses of the image will use the copy. If the image is ever
used off a GrDirectContext we fall over to the copy. If the copy is
never used and never can be used by the next flush then the render
task that populates it is marked "skipped" and we don't perform the
copy.
Bug: skia:11208
Change-Id: Id255f4a733acc608c8a53c1a5633207aeafc404b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/366282
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This is a reland of 40a9061203
Original change's description:
> rename GrSDFTOptions to GrSDFTControl
>
> Change-Id: Ie03fce7a99a9f71b18d54e3cd35e7675fb7f8912
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379616
> Commit-Queue: Herb Derby <herb@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
Change-Id: Iba0816159a8ef99448a0040c487e56700b96be5d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379845
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Herb Derby <herb@google.com>
This performs the same simplifications as the control-flow phase, but at
IR generation time. There's no visible difference in the tests (besides
DSL) because these aren't new optimizations; they're just happening
at a different phase of compilation.
Change-Id: I26d241167b0e690b23f8f4370339714783c8d6fd
Bug: skia:11343
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/371482
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>
This check now runs at function finalization time, before constant
propagation has occurred; this affected the "DeadIfStatement" test.
Our detection isn't smart enough to realize that a loop will run zero
times, so it treats `for` and `while` loops as always running at least
once. This isn't strictly correct, but it actually mirrors how the CFG
implementation works anyway. The only downside is that we would not flag
code like `for (i=0; i<0; ++i) { return x; }` as an error.
Change-Id: I5e43a6ee3a3993045559f0fb0646d36112543a94
Bug: skia:11377
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379056
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
In addition to the unsurprising changes to eliminate references to
src/, we also had to tighten up some C++17-isms as they are not
permitted in public headers.
Change-Id: Ie5005a33d7a135e69fb66beca5e7a5f960dbd453
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378496
Reviewed-by: Brian Osman <brianosman@google.com>
PictureShader = picture + tiling + depth/colorspace + filtering [+ scale]
Today we cache the imageshader that is used to rendering. However, the
key for that cache is the pictureshader's ID itself... which means if
we have several, all using the same picture (but maybe diff tiling) we
would create dup cache entries.
Idea:
1. only cache the image (rastered picture), not an imageShader
2. key the cache on the picture's ID, not the shader's
Several implications of this:
1. Should get more cache reuse, since we don't care about the
shader's ID (which is just wrapping a picture+tiling, etc.)
2. We also eliminate the indirection of creating a PictureImage. Instead
we're creating real (pixel) images, and caching those. This removes one
extra layer of "cache".
Idea: when we cache something for pict
Change-Id: I51cf4e9bff3c91ce1872876597d3d565039d8c7a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/377844
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This should be legal, and we support this, but some versions of Android
do not: http://screen/3bkQewHF3xUMn5v There's no point in allowing
these shaders to exist; they can't compile on real-world clients, and
these vardecls are borderline meaningless (as the variables being
declared aren't reachable by any other statements).
Change-Id: Ie1351933c90caee9124eeab8983364ec030b2653
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379584
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This reverts commit 50b1b2b90d.
Reason for revert: ending experiment
Original change's description:
> Disable control-flow analysis in SkSL. (Performance experiment)
>
> This CL will be used to test for potential performance regressions (or
> improvements?) that we might incur by disabling this optimization pass.
>
> It will be reverted in ~1 day.
>
> Change-Id: I775cdb0c95df81fa25ebbd66e4ff01f64c660f68
> Bug: skia:11319
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/378456
> Commit-Queue: John Stiles <johnstiles@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: Ie385a82db237ff5651348d82b9651f8ba09375b9
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379581
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Covers some common geometry processors, texture effect, etc.
This also rearranges how fp keys are arranged in the overall key. We no
longer include the key size as part of the key - this made no sense.
Instead, we explicitly include the number of children. We also put all
data for one fp before any children, so the tree can be reconstructed
more-or-less top-down.
Finally, added an "addBool" helper that reads nicer than addBits(1)
everywhere.
Bug: skia:11372
Change-Id: I4e35257fb5923d88fe6d7522109a0b3f4c4017d4
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379059
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
