This is a reland of 91c1d08bc3
Original change's description:
> SkSL is now pickier about type conversions
>
> Bug: skia:
> Change-Id: I4e8b8f229f4e4344f160b0dbb41832764d0b75bd
> Reviewed-on: https://skia-review.googlesource.com/c/188311
> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
Bug: skia:
Change-Id: I727cad061afc0a5ee6f4d2df789330d809dd110a
Reviewed-on: https://skia-review.googlesource.com/c/189643
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
All uses have been converted to SkPMColor4f (or similar).
Bug: skia:
Change-Id: I220bd5eaf6c35b17321c1e8bc92ace7ff92908c4
Reviewed-on: https://skia-review.googlesource.com/c/162749
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Getters for GrFragmentProcessor's optimization flags and state are
exposed as constant fields on the fragmentProcessor variable in sksl.
The CPP code generation then emits extra instructions to invoke the
particular getter when building the final sksl string at runtime.
Bug: skia:
Change-Id: If6bf4f4df6c446fb6331484d36effb1386172918
Reviewed-on: https://skia-review.googlesource.com/152381
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This reverts commit d0440195d5.
Reason for revert: Fixes size_t -> int that was triggering ASAN failures.
Original change's description:
> Revert "Redo how extra emit code flushing operates"
>
> This reverts commit 9b8181b05a.
>
> Reason for revert: <INSERT REASONING HERE>
>
> Original change's description:
> > Redo how extra emit code flushing operates
> >
> > The previous implementation of flushEmittedCode(), that flushed on
> > demand when a process() was encountered, was brittle and susceptible to
> > mangling the expected sksl when fOut was modified outside of its
> > control. Given that writeFunction() and generateCode() in the parent
> > class all do this, it's possible to generate a simple SkSL snippet that
> > would generate a CPP file that builds invalid final SkSL:
> >
> > ```
> > in fragmentProcessor child;
> > bool someGlobalVar = ...;
> > void main() {
> > if (someGlobalVar) {
> > sk_OutColor = process(child, sk_InColor);
> > } else {
> > sk_OutColor = half4(1);
> > }
> > }
> > ```
> >
> > The CPP generated code *should* insert 'bool someGlobalVar' at the start
> > but because of the early flush from the child process and because of
> > how fOut was overwritten, someGlobalVar's declaration is put into a
> > stream that is not visible to the flush and ends up being inserted into
> > the output sksl in an incorrect location (namely after the if condition
> > that depends on it).
> >
> > This CL updates the extra emitted code logic to support multiple blocks
> > of extra CPP code. When a flush point occurs in SkSL writing, a special
> > token is inserted into the SkSL and a new CPP code buffer is associated
> > with that token. Then once all of the SkSL is accumulated into the root
> > output stream, it is processed into sections for each extra CPP block.
> > Special logic is done so that the SkSL that is emitted before the next
> > CPP block terminates at the end of the last valid statement before the
> > special token.
> >
> > A unit test demonstrating this failure condition is added to SkSLFPTest
> > and the CL properly passes. Since this bug did not trigger on existing
> > .fp files, the updated generator does not modify the generated FPs.
> >
> > Bug: skia:
> > Change-Id: Ib74911942080f1b964159807a06805bc52898789
> > Reviewed-on: https://skia-review.googlesource.com/152321
> > Commit-Queue: Michael Ludwig <michaelludwig@google.com>
> > Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
>
> TBR=ethannicholas@google.com,michaelludwig@google.com
>
> Change-Id: Id0f908453b596873f43b86a1c14eed48b2474a76
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:
> Reviewed-on: https://skia-review.googlesource.com/152660
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
> Commit-Queue: Michael Ludwig <michaelludwig@google.com>
TBR=ethannicholas@google.com,michaelludwig@google.com
Change-Id: I3ccf2fee6ef96c6102dbe1c2c2ef6c14a701b8fd
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/152663
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
This reverts commit 9b8181b05a.
Reason for revert: <INSERT REASONING HERE>
Original change's description:
> Redo how extra emit code flushing operates
>
> The previous implementation of flushEmittedCode(), that flushed on
> demand when a process() was encountered, was brittle and susceptible to
> mangling the expected sksl when fOut was modified outside of its
> control. Given that writeFunction() and generateCode() in the parent
> class all do this, it's possible to generate a simple SkSL snippet that
> would generate a CPP file that builds invalid final SkSL:
>
> ```
> in fragmentProcessor child;
> bool someGlobalVar = ...;
> void main() {
> if (someGlobalVar) {
> sk_OutColor = process(child, sk_InColor);
> } else {
> sk_OutColor = half4(1);
> }
> }
> ```
>
> The CPP generated code *should* insert 'bool someGlobalVar' at the start
> but because of the early flush from the child process and because of
> how fOut was overwritten, someGlobalVar's declaration is put into a
> stream that is not visible to the flush and ends up being inserted into
> the output sksl in an incorrect location (namely after the if condition
> that depends on it).
>
> This CL updates the extra emitted code logic to support multiple blocks
> of extra CPP code. When a flush point occurs in SkSL writing, a special
> token is inserted into the SkSL and a new CPP code buffer is associated
> with that token. Then once all of the SkSL is accumulated into the root
> output stream, it is processed into sections for each extra CPP block.
> Special logic is done so that the SkSL that is emitted before the next
> CPP block terminates at the end of the last valid statement before the
> special token.
>
> A unit test demonstrating this failure condition is added to SkSLFPTest
> and the CL properly passes. Since this bug did not trigger on existing
> .fp files, the updated generator does not modify the generated FPs.
>
> Bug: skia:
> Change-Id: Ib74911942080f1b964159807a06805bc52898789
> Reviewed-on: https://skia-review.googlesource.com/152321
> Commit-Queue: Michael Ludwig <michaelludwig@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
TBR=ethannicholas@google.com,michaelludwig@google.com
Change-Id: Id0f908453b596873f43b86a1c14eed48b2474a76
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/152660
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
The previous implementation of flushEmittedCode(), that flushed on
demand when a process() was encountered, was brittle and susceptible to
mangling the expected sksl when fOut was modified outside of its
control. Given that writeFunction() and generateCode() in the parent
class all do this, it's possible to generate a simple SkSL snippet that
would generate a CPP file that builds invalid final SkSL:
```
in fragmentProcessor child;
bool someGlobalVar = ...;
void main() {
if (someGlobalVar) {
sk_OutColor = process(child, sk_InColor);
} else {
sk_OutColor = half4(1);
}
}
```
The CPP generated code *should* insert 'bool someGlobalVar' at the start
but because of the early flush from the child process and because of
how fOut was overwritten, someGlobalVar's declaration is put into a
stream that is not visible to the flush and ends up being inserted into
the output sksl in an incorrect location (namely after the if condition
that depends on it).
This CL updates the extra emitted code logic to support multiple blocks
of extra CPP code. When a flush point occurs in SkSL writing, a special
token is inserted into the SkSL and a new CPP code buffer is associated
with that token. Then once all of the SkSL is accumulated into the root
output stream, it is processed into sections for each extra CPP block.
Special logic is done so that the SkSL that is emitted before the next
CPP block terminates at the end of the last valid statement before the
special token.
A unit test demonstrating this failure condition is added to SkSLFPTest
and the CL properly passes. Since this bug did not trigger on existing
.fp files, the updated generator does not modify the generated FPs.
Bug: skia:
Change-Id: Ib74911942080f1b964159807a06805bc52898789
Reviewed-on: https://skia-review.googlesource.com/152321
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Easily supports mapping ctypes to sksl types with templates that
specify how to send data to the GPU and how to track state changes.
The template logic and type mappings are defined in
SkSLCPPUniformCTypes.* while SkSLCPPCodeGenerator is updated to
utilize it.
It also updates the supported ctypes to properly generate code for
SkPoint, SkIPoint, SkIRect, and GrColor4f. The code generated for
'in uniforms' now also correctly supports conditional uniforms.
Bug: skia:
Change-Id: Ib7c0a873bdd68a966b6a00871f33102dfa2c432d
Reviewed-on: https://skia-review.googlesource.com/150129
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
--
This expands sksl's capabilities with .fp files. Previously, it was possible to declare "in fragmentProcessor foo" and emit it automatically when "process(foo);" was called. This adds a variant of process that takes a second argument, which must be a half4 expression. This argument specifies the value, or dynamic expression calculated earlier in the parent shader, to use as sk_InColor by the child.
The CL is longer than anticipated because of properly handling dependencies between previous sksl statements and the input to the child. The original writeEmitCode() collected all extra emission code (the calls to build->emitChild) and put them before any call to codeAppendf. This makes it impossible to use a parent's variable, or the output of another child, as the input for process.
To solve this, there is now a flushEmittedCode() function that takes over the logic of outputting the extra emission code and the necessary codeAppendf calls. When invoked, it (by default) only appends completed sksl statements, and places any current expression back at the beginning of the output stream. It now updates fFormatArgs and fExtraEmitCodeCode as it consumes their contents. This allows writeFunctionCall() for a call to "process" to flush all previous statements before it adds its emit child code.
Bug: skia:
Change-Id: I63c41af6f3e0620aa890d10d14436ee6244f0051
Reviewed-on: https://skia-review.googlesource.com/148395
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@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>
Bug: skia:
Change-Id: Ia50c1f750e4626211f012ae7543db126b10134c3
Reviewed-on: https://skia-review.googlesource.com/61906
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-allocated IDs mean that the IDs depend upon the order in which
classes happen to get initialized and are therefore not consistent
from run to run. This change paves the way for a persistent shader
cache by fixing the IDs in an enum.
Bug: skia:
Change-Id: I3e923c6c54f41b3b3eb616458abee83e0909c09f
Reviewed-on: https://skia-review.googlesource.com/56401
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This reverts commit 05d5a13fea.
Reason for revert: looks like it broke filterfastbounds
Original change's description:
> Revert "Revert "Switched highp float to highfloat and mediump float to half.""
>
> This reverts commit 1d816b92bb.
>
> Bug: skia:
> Change-Id: I388b5e5e9bf619db48297a80c9a80c039f26c9f1
> Reviewed-on: https://skia-review.googlesource.com/46464
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
TBR=bsalomon@google.com,ethannicholas@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: skia:
Change-Id: Iddf6aef2ab084aa73da7ceebdfc303a1d2b80cde
Reviewed-on: https://skia-review.googlesource.com/47441
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
The ultimate goal is to end up with "float" and "half", but this
intermediate step uses "highfloat" so that it is clear if I missed a
"float" somewhere. Once this lands, a subsequent CL will switch all
"highfloats" back to "floats".
Bug: skia:
Change-Id: Ia13225c7a0a0a2901e07665891c473d2500ddcca
Reviewed-on: https://skia-review.googlesource.com/31000
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Bug: skia:
Change-Id: Ifbccc6d6880522177ac7a0ae2183be64a3ebfe50
Reviewed-on: https://skia-review.googlesource.com/34681
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
On desktop, this saves just over 5% of the time in the SkSL compiler.
As written, the code will now build either way, so it's much easier to
switch back (or even have some platforms use SkString, if that's ever
required).
Bug: skia:
Change-Id: I634f26a4f6fcb404e59bda6a5c6a21a9c6d73c0b
Reviewed-on: https://skia-review.googlesource.com/34381
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Change-Id: I985e54a071338e99292a5aa2f42c92bc115b4008
Reviewed-on: https://skia-review.googlesource.com/32760
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Fixes build of Skia lib when GR_TEST_UTILS=0
Makes GR_TEST_UTILS=0 for official builds
Makes "Mini" builder bot exercise building GPU with is_official_build=true
Bug: skia:6786
Change-Id: I6186683a3a216d2e779645bd9e8276a66bcff4d5
Reviewed-on: https://skia-review.googlesource.com/21524
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:
Change-Id: I641b206fc3bc19ac190ad94ee755ab9e1caab9b3
Reviewed-on: https://skia-review.googlesource.com/21341
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
This reverts commit c070939fd1.
Reason for revert:
This has some knock-on effects in the generation of Android.bp from our GN files. See gn/gn_to_bp.py? We're seeing things like "tmp/tmpsBVycx/gen/" end up in the include search path in Android.bp, which obviously don't exist there...
Original change's description:
> Re-land sksl fragment processor support
>
> This reverts commit ed50200682.
>
> Bug: skia:
> Change-Id: I9caa7454b391450620d6989dc472abb3cf7a2cab
> Reviewed-on: https://skia-review.googlesource.com/20965
> Reviewed-by: Ben Wagner <benjaminwagner@google.com>
> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
TBR=benjaminwagner@google.com,ethannicholas@google.com
Change-Id: I502486b5405923b322429219f4cc396a45a14cea
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/20990
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Bug: skia:
Change-Id: Ia3b0305c2b0c78074303831f628fb01852b90d34
Reviewed-on: https://skia-review.googlesource.com/17843
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
