This will allow us to reuse this logic in Graphite.
Change-Id: I649dcd3893a1355af457a2583a6db3066fb87c9a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/532758
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Previously, we used unscoped blocks for two similar functions:
- Rewrite one statement as two simpler statements:
`int a, b;` -> `int a; int b;`
- Group together multiple statements without braces. e.g. the inliner
uses unscoped Blocks to rearrange statements.
Conceptually, these are different from the debugger's perspective. The
compound statements should be treated as one unit; the grouped
statements should be treated individually (and the enclosing Block
should be ignored). A Block now contains a BlockKind enum to
distinguish between these cases.
Change-Id: Ie14a570bb46992689fb96b8fd3b67f2ca6e5239f
Bug: skia:13189
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/528655
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This adopts a trick from SkVM to avoid sorting entirely.
Change-Id: I586c8a3613b48241842a7d8eba1c9d68a4717f83
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/528368
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Caveat: on my machine, Nanobench doesn't detect any change (pro or con)
on this CL.
I'm working under the assumption that function calls have a non-zero
cost--they may be inlined (bloating code size), or not (incurring the
costs of a function call, register push/popping, etc). This CL avoids
making six calls to $blend_set_color_saturation by using two half3
variables. These half3s are used to swizzle the result--they contain two
zeros and a one, so multiplying them by a scalar will put the result in
the desired component. I've also made some very minor simplifications to
the math that were made possible by reordering.
Change-Id: I0c1ef88d165365376078846324be8bb723548512
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/528043
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Today I learned that `mix(a, b, 1)` can reduce precision. Ternaries do
not suffer from this problem.
Change-Id: I58814d00193ccbff53960030d163d31c49234f6c
Bug: skia:9320
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/528161
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
The inliner can do a better job with functions that only have a single
return by eliding a temp variable. In this case, it was simple to adapt.
Change-Id: I9a5ee26cf546db1b2647cdf95d4cdba6649ea19b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/528160
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This reverts commit 2e6f60f423.
Reason for revert: draws black incorrectly in various iPhone 8 tests
Original change's description:
> Fix color fringes on blend_hue and blend_saturation.
>
> Previously, we checked for division against zero, but didn't do anything
> to prevent division against extraordinarily small values. Now, we only
> saturate if the delta between max and min is greater than 0.00001.
>
> Change-Id: I7d1df3430941c7e1a7f94e597d5449f9259612d6
> Bug: skia:9320
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527498
> Auto-Submit: John Stiles <johnstiles@google.com>
> Reviewed-by: Greg Daniel <egdaniel@google.com>
> Commit-Queue: Greg Daniel <egdaniel@google.com>
Bug: skia:9320
Change-Id: Id83376080eed684577b3592c5e1bee3c80fc3fc9
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/528038
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Derek Sollenberger <djsollen@google.com>
We use multiplication by 1 or -1 to branchlessly choose one of `min` and
`max` in the same function.
Change-Id: I44cf747feeae75a9c3e00f36e112e0a429871e86
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527596
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Hard-light is just overlay with the parameters reversed.
Change-Id: I6cf5963b1252cba3a7b71a56f4094a070188f8b2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527503
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
These functions were functionally almost identical, except:
- Sometimes sda/dsa are flipped
- Sometimes the saturation is not updated
We now have one method (blend_hslc) which can do all four blend
operations. It takes two new parameters ("flip" and "saturate") to
handle these four variations.
This reduces our shader count on some of our most shader-heavy slides
(e.g. aaxfermodes, xfermodeimagefilter) at a pretty reasonable cost.
Change-Id: Ifa8a48399851a9badb5d50038de1e25e60d44ebd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527281
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Previously, we checked for division against zero, but didn't do anything
to prevent division against extraordinarily small values. Now, we only
saturate if the delta between max and min is greater than 0.00001.
Change-Id: I7d1df3430941c7e1a7f94e597d5449f9259612d6
Bug: skia:9320
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527498
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Requires tweaking one inliner test to avoid an Intel driver bug (on
ANGLE).
Bug: chromium:709351
Cq-Do-Not-Cancel-Tryjobs: true
Change-Id: I08fac938396d6b90805ba9650c7a520af888bc12
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/503819
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Test "InlinerHonorsGLSLOutParamSemantics" was failing on Wembley devices
and is now disabled on that GPU.
Also, it turns out that the inliner has ignored functions with out
params for a long time now, but our test names haven't been updated to
account for this. So, did some additional cleanup:
- "InlinerHonorsGLSLOutParamSemantics" (the test in question) has been
moved to shared/ and renamed to "OutParamsAreDistinct."
- Removed test "OutParamsNoInline" as it is functionally the same as
"OutParams".
Change-Id: I1431ed197b9216cb482eee4f5e4eb2579a5303f7
Bug: skia:12858
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/502303
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
The test has been moved to shared/, since it's a valid test, but it is
no longer related to inlining, as the inliner no longer attempts to
inline functions with inouts at all.
Also, one function here (outParameterIgnore) actually invoked undefined
behavior and has been removed. According to the GLSL ES2 docs: "If a
function does not write to an out parameter, the value of the actual
parameter is undefined when the function returns." SkVM leaves the value
unchanged, so SKSL_TEST_CPU would pass, but a GPU might clear it (and in
fact, my GPU does).
Change-Id: I77c77ed1354bc980344ec5c406992bd62015f5e5
Bug: skia:11919
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/499752
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This guards against any potential for conflict with user code.
Change-Id: Iecaf3ead5f8ada50b6dc159a4ad9e7f3e371edc7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/499296
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This eliminates a handful of useless statements. They were harmless, but
presumably this saves optimization work for the GLSL compiler.
(Patterned after http://review.skia.org/496377 )
Change-Id: Ibe135750488a9917b982dcac67f22b3765412ee1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/496599
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Our inliner would ignore any functions with `inout` parameters, because
inlining them properly was more complex than just leaving the function
call. However, real-world code can sometimes contain helper functions
that have `inout` params that are never used at all (e.g. an uber-shader
with some features turned off).
We now read the ProgramUsage and check to see whether or not the
`inout`-qualified parameter is actually modified. If it's never changed,
the function now remains a candidate for inlining.
Change-Id: I92e494f94cc070801cb9aa28bd13faa689b806b6
Bug: skia:12636
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/470299
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Functions which don't write to their out params should be safe to
inline, but we currently don't recognize this.
Change-Id: I753e48067c7be4473675ef6c95e61af17dc5ae41
Bug: skia:12636
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/470298
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Intrinsic-call optimization can be triggered during inlining. In this
case, inlining turned `normalize(x)` into `normalize(constant-value)`.
DSL is used to implement optimizations for a handful of intrinsic calls,
including `normalize`, which internally relies on `length`.
The DSL expects that it can use the IRGenerator to handle function
calls. This was not working because we were finished with the initial
compilation pass, and the IRGenerator's symbol table is removed when
finish() was called.
We now temporarily give a symbol table back to the IRGenerator while
the inliner runs. We remove it again as soon as inlining is complete.
Change-Id: I6da98788d93749ffeb008c1f4c3f72b436e8ceeb
Bug: oss-fuzz:37994
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/445956
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Bug: skia:11296
Change-Id: I7d41614957d6fa535faadebbeca890b54b6977ac
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/425996
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
This is a reland of c6260f9742
Problematic DeadReturn.sksl test cases have been moved to DeadReturnES3.
Original change's description:
> Eliminate unreachable code during optimization.
>
> The Adreno 5xx and 6xx previously failed the SkSLStaticSwitchInline
> test; the driver struggled to interpret code with multiple return
> statements in a row. We now detect unreachable statements and eliminate
> them.
>
> Change-Id: I344d632f2488ca65b0635b37bebffe6e4fb607c5
> Bug: skia:12012
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/410256
> Auto-Submit: John Stiles <johnstiles@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
Bug: skia:12012
Change-Id: I748e8761cbc71c811b5ad8fe49186f980261d8b9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/410793
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 reverts commit c6260f9742.
Reason for revert: Nexus 7 is grumpy, fails SkSLDeadReturn_GPU
https://ci.chromium.org/raw/build/logs.chromium.org/skia/539fff50ff093c11/+/annotations
Original change's description:
> Eliminate unreachable code during optimization.
>
> The Adreno 5xx and 6xx previously failed the SkSLStaticSwitchInline
> test; the driver struggled to interpret code with multiple return
> statements in a row. We now detect unreachable statements and eliminate
> them.
>
> Change-Id: I344d632f2488ca65b0635b37bebffe6e4fb607c5
> Bug: skia:12012
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/410256
> Auto-Submit: John Stiles <johnstiles@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: Ia3db1f1b28417e479e2d71a4a6ed94a007e47cf9
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:12012
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/410780
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
The Adreno 5xx and 6xx previously failed the SkSLStaticSwitchInline
test; the driver struggled to interpret code with multiple return
statements in a row. We now detect unreachable statements and eliminate
them.
Change-Id: I344d632f2488ca65b0635b37bebffe6e4fb607c5
Bug: skia:12012
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/410256
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
In particular, this optimizes abs() and sign() when all inputs are known
at compile time. This resolves a TODO on a test case in
`IllegalIndexing.rts`.
Change-Id: Ica310522a85b42dc7ae255bd25004a6629d04176
Bug: skia:10835
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/405676
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This CL adds a RuntimeEffect option flag which skips over the various
`strictES2Mode` checks sprinkled throughout IR generation.
Runtime Effects still won't allow a lot of ES3 things (the Pipeline
stage will reject unsupported statement types, SkVM doesn't support most
non-ES2 constructs, etc). However, this change will give us the ability
to test many more features involving arrays and structs that previously
were off-limits due to ES2 restrictions, and will shore up some
legitimate gaps in our testing. This is a useful starting point to allow
for improved test coverage.
Change-Id: I4a5bc43914e65fc7e59f1cecb76a0ec5a7f05f2f
Bug: skia:11209
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/402157
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>
Previously, these tests were never actually executed, only read during
code review. They are now properly tested for correctness whenever dm
is run. Non-ES2 compliant statements (do/while/switch) are unfortunately
excluded here, as they are not compatible with Runtime Effects yet.
Change-Id: I965c782baad6f8dd3961a400ae791fb2c1f844d3
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/389296
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This will be implemented in Metal and SPIR-V in followup CLs.
Change-Id: I397b4db40b15dd54cf1d8a17f414c3fe184b48d2
Bug: skia:10851
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/387638
Auto-Submit: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This will allow the inliner to successfully do more work in a single
pass.
Change-Id: I26e8831737c10bdf9a35eebd94ea8b74f6487077
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/386916
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Arguments without side-effects that aren't read from more than once can
be moved directly into the inlined function, and don't need a scratch
variable. This can allow functions like `guarded_divide` to inline
completely in more cases.
Change-Id: I0bfce35635cf9779f4af1bc0790da966ccfe4230
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/386678
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
It is difficult to do this both efficiently and correctly while honoring
GLSL semantics (which require the lvalues to be kept distinct, even when
they point to the same variable). We could make it work by making copies
of every out parameter in each direction (going in for inouts, and
coming out for outs and inouts).
However, this could be self-defeating if it makes it harder for the
driver to track variable lifetimes. Simply opting out of inlining these
functions entirely seems like the best tradeoff; let the driver optimize
them if it can, and we can enjoy reduced complexity in the SkSL inliner.
Change-Id: I62f7b4550cc181cfe789e4f2ff4e408ba1baf9cb
Bug: skia:11326
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/370257
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This allows us to remove 100 LOC from the inliner and is very unlikely
to affect any existing benchmark. We don't have any evidence to support
the idea that a one-iteration `for` loop with `continue`-based exits
will be any faster than a standard function call on any existing GPU.
Our fragment processors are generally written to avoid early returns,
in large part to avoid hitting this path.
This drastically impacts BlendEnum.sksl (which can no longer flatten out
a switch over every blend function in SkSL) but is otherwise a wash.
See: http://go/optimization-in-sksl-inliner suggestion 4(a)
Change-Id: I1f9c27bcd7a8de46cc4e8d0b9768d75957cf1c50
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385377
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This can eliminate const variables which have been completely folded
away, unnecessary synthetic variables created during codegen/inlining,
or code that simply didn't need to exist at all.
Change-Id: I37a65e455e6527a6a6c2f4dde918f48d84dc2638
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/383496
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
This reverts commit 92748af1a5.
Reason for revert: SkSLCommaSideEffects_GPU crashing on Android
Original change's description:
> Inline functions of the form 'return (expr)' only.
>
> 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>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: I6a670dacaa58fe3386ff50375ac6d1cac4fd7f2c
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385161
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
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>
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>
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>
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>
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>
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>
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>
This reverts commit e4da7b672f.
Reason for revert: breaks SkSLBench perf test
Original change's description:
> Migrate if-statement simplifyStatement logic to IfStatement::Make.
>
> This performs essentially the same simplifications as before, just at
> a different phase of compilation.
>
> Change-Id: Ia88df6857d4089962505cd1281798fda74fd0b02
> Bug: skia:11343, skia:11319
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376177
> Commit-Queue: John Stiles <johnstiles@google.com>
> Auto-Submit: John Stiles <johnstiles@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com
Change-Id: I0051188ffe69426904066eb60a932435efdc2af8
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:11343
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/379062
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This performs essentially the same simplifications as before, just at
a different phase of compilation.
Change-Id: Ia88df6857d4089962505cd1281798fda74fd0b02
Bug: skia:11343, skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/376177
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
These variables were later being eliminated by the dead-code-elimination
pass, so you can't see them directly in the final output, but removing
them affects the name mangling off all future symbols, so it causes an
enormous ripple effect in the diff. And of course, it's a waste of time
and memory to synthesize IRNodes just to destroy them later.
If we disable control-flow analysis, we lose the dead-code-elimination
pass entirely; this change is also beneficial for emitting better code
when optimizations are turned off.
Change-Id: I882b3be4f3fd99b77d99b6abe128f26bb9252c89
Bug: skia:11319
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/375776
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
The optimizer now properly recognizes all types of exits from a switch
statement. Break, continue and return are all potential exits and need
to be considered when determining the exit path from the switch.
Previously, dead code elimination was hiding the effects of this bug
from us, but it meant that an optimized switch had the potential to
generate lots of worthless IR nodes which then needed to be detected and
eliminated by the CFG. In particular, this affected the enum form of
blend, causing a catastrophic amount of extra work to be done.
Change-Id: If857e38cadfc016884624ea4db25a273ad3dce5b
Bug: skia:11352
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/372958
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
When we detect a static switch, the optimizer finds the matching switch-
case and eliminates all the other switch-cases. It handles case
fall-through by scanning forward and looking for an unconditional break.
However, the inliner has an interesting quirk--it can replace `return`
statements inside of a switch with `continue` statements, since the body
of the inlined function has been wrapped with a for-loop to allow for
early exits. The optimizer does not recognize these continue statements
as exits from the switch (although they certainly qualify), so it
treats continues as fallen-through and keeps emitting switch-cases.
The dead-code elimination pass was actually doing us a favor here and
eliminating the excess code later. A flag was added to disable DCE in
order to reveal the problem in a test.
Change-Id: I8ff19fde5e32d0ab73d7c5411da40cb953a446f5
Bug: skia:11352
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/372956
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
GLSL ES2 documentation on out parameters: "Evaluation of an out
parameter results in an l-value that is used to copy out a value when
the function returns."
The inliner does not do any alias checking when inlining an `out` param.
That is, passing the same variable to two separate `out` parameters
would not generate two distinct lvalues in the inlined code; it reuses
the same variable for each out-params in the inlined code.
(Amusingly, our CFG can fully optimize away this test code so it just
returns "red".)
Change-Id: Ib781d2cfdac54f01b6abe159af0c84ff24ff6976
Bug: skia:11326
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/370256
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Uses the pipeline-stage callback mechanism. It mangles the type name
(with a test to verify that this works), and then calls defineStruct
with the entire SkSL struct definition string.
Bug: skia:10939
Change-Id: If14cf1b11faaa80ad8d4086cdacf68532bac43fc
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368809
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Fixes another instance of anglebug.com/2098 with advanced blend
functions.
Change-Id: I91863723d8b4c33ab2f5a527fe0374e8947bba16
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/368813
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This creates a helper function, _entrypoint, which invokes main() and
assigns its result into sk_FragColor. We also make sure to prevent
sk_FragColor from being dead-stripped from the code during IR
generation.
At present this is useful for allowing our SkSL test shaders to compile.
Change-Id: I2d7fab0e1959a77778ffdb18ca569e869bcaeece
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/358525
Commit-Queue: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
