AuroraMiddleware/skia2
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
b2b1d6553e
skia2/tests/sksl/blend/BlendSoftLight.asm.frag

453 lines
15 KiB
GLSL
Raw Normal View History

Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpCapability Shader
%1 = OpExtInstImport "GLSL.std.450"
OpMemoryModel Logical GLSL450
Change sksl tests to avoid SPIR-V validation errors 'in' variables without locations aren't allowed. Use uniforms instead. Bug: skia:11738 Change-Id: Ic066106deb7409cff154b4be7cfb3e03a7025c7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385000 Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: John Stiles <johnstiles@google.com>
2021-03-15 18:58:57 +00:00
OpEntryPoint Fragment %main "main" %sk_FragColor %sk_Clockwise
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpExecutionMode %main OriginUpperLeft
OpName %sk_FragColor "sk_FragColor"
OpName %sk_Clockwise "sk_Clockwise"
Change sksl tests to avoid SPIR-V validation errors 'in' variables without locations aren't allowed. Use uniforms instead. Bug: skia:11738 Change-Id: Ic066106deb7409cff154b4be7cfb3e03a7025c7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385000 Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: John Stiles <johnstiles@google.com>
2021-03-15 18:58:57 +00:00
OpName %_UniformBuffer "_UniformBuffer"
OpMemberName %_UniformBuffer 0 "src"
OpMemberName %_UniformBuffer 1 "dst"
Use $ prefix on built-in private functions. 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>
2022-01-25 14:42:12 +00:00
OpName %soft_light_component_Qhh2h2 "soft_light_component_Qhh2h2"
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpName %DSqd "DSqd"
OpName %DCub "DCub"
OpName %DaSqd "DaSqd"
OpName %DaCub "DaCub"
OpName %main "main"
OpDecorate %sk_FragColor RelaxedPrecision
OpDecorate %sk_FragColor Location 0
OpDecorate %sk_FragColor Index 0
OpDecorate %sk_Clockwise BuiltIn FrontFacing
Change sksl tests to avoid SPIR-V validation errors 'in' variables without locations aren't allowed. Use uniforms instead. Bug: skia:11738 Change-Id: Ic066106deb7409cff154b4be7cfb3e03a7025c7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385000 Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: John Stiles <johnstiles@google.com>
2021-03-15 18:58:57 +00:00
OpMemberDecorate %_UniformBuffer 0 Offset 0
OpMemberDecorate %_UniformBuffer 0 RelaxedPrecision
OpMemberDecorate %_UniformBuffer 1 Offset 16
OpMemberDecorate %_UniformBuffer 1 RelaxedPrecision
OpDecorate %_UniformBuffer Block
OpDecorate %11 Binding 0
OpDecorate %11 DescriptorSet 0
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %21 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %22 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %23 RelaxedPrecision
OpDecorate %24 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %25 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %30 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %31 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %32 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %33 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %34 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %35 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %36 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %37 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %38 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %39 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %40 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %41 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %42 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %43 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %44 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %46 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %47 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %48 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %49 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %50 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %51 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %52 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %53 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %54 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %55 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %56 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %57 RelaxedPrecision
Revert "Refactored SPIR-V RelaxedPrecision handling" This reverts commit 9e476b798f03930b6bc7fb6eacab574f197e172d. Reason for revert: Angry Vulkan bots Original change's description: > Refactored SPIR-V RelaxedPrecision handling > > The RelaxedPrecision decoration is now handled by nextId(), to make it > easier to see all spots where a RelaxedPrecision decoration might be > necessary. The goal of this initial refactor is not to actually fix the > issues with RelaxedPrecision decorations, but rather to lay the > groundwork for doing so in followup CLs. > > The initial intent of this change was to not affect the SPIR-V at all, > saving modifications for followups, but there ended up being three kinds > of changes to the output: > > 1. Doing things at nextId() time rather than later means some > decorations move to an earlier spot in the output. This results in > diffs, but should not cause any behavioral changes. > 2. We were incorrectly tagging bools as RelaxedPrecision in some > situations. By funneling things through fewer code paths, the refactor > would have caused this to happen in even more situations, and the code > responsible for the bug was being rewritten in this CL anyway, so it > seemed worth just fixing the issue as part of this change. > 3. Funneling things through fewer code paths ended up adding > (correct) RelaxedPrecision modifiers to binary operations that had > previously been missing them. It seemed better to just let it happen > than to try to maintain bug-for-bug compatibility with the previous > approach. > > Change-Id: Ia9654d6b5754e2c797e02226660cb618c9189b36 > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/384318 > Commit-Queue: Ethan Nicholas <ethannicholas@google.com> > Reviewed-by: Brian Osman <brianosman@google.com> TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com Change-Id: I9ada728e5fd5798bc1179640560c2e6045b7efd1 No-Presubmit: true No-Tree-Checks: true No-Try: true Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385158 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-15 21:24:32 +00:00
OpDecorate %58 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %59 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %60 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %61 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %62 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %63 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %64 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %66 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %67 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %68 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %69 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %70 RelaxedPrecision
OpDecorate %DSqd RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %77 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %78 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %79 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %80 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %81 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %DCub RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %83 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %84 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %85 RelaxedPrecision
OpDecorate %DaSqd RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %87 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %88 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %89 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %90 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %91 RelaxedPrecision
OpDecorate %DaCub RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %93 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %94 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %95 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %96 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %97 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %98 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %99 RelaxedPrecision
OpDecorate %101 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %102 RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %103 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %105 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %106 RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %107 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %108 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %109 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %110 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %111 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %112 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %114 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %115 RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %116 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %117 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %118 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %119 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %120 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %121 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %122 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %123 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %124 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %125 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %127 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %128 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %129 RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %130 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %131 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %132 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %133 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %134 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %135 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %136 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %137 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %138 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %139 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %140 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %141 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %142 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %143 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %144 RelaxedPrecision
OpDecorate %145 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %146 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %147 RelaxedPrecision
OpDecorate %148 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %149 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %150 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %151 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %152 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %153 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %154 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %155 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %156 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %157 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %158 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %159 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %160 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %161 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %162 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %163 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %164 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %165 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpDecorate %166 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %167 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %168 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %169 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %170 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %171 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %172 RelaxedPrecision
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
OpDecorate %173 RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %181 RelaxedPrecision
OpDecorate %182 RelaxedPrecision
OpDecorate %192 RelaxedPrecision
OpDecorate %194 RelaxedPrecision
OpDecorate %195 RelaxedPrecision
OpDecorate %198 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %199 RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %203 RelaxedPrecision
OpDecorate %204 RelaxedPrecision
OpDecorate %207 RelaxedPrecision
OpDecorate %208 RelaxedPrecision
OpDecorate %212 RelaxedPrecision
OpDecorate %213 RelaxedPrecision
OpDecorate %216 RelaxedPrecision
OpDecorate %217 RelaxedPrecision
OpDecorate %221 RelaxedPrecision
OpDecorate %222 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %224 RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %225 RelaxedPrecision
OpDecorate %226 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %228 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %229 RelaxedPrecision
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
OpDecorate %230 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %231 RelaxedPrecision
Revert "Revert "Added more RelaxedPrecision decorations"" This reverts commit 7bba1f55e85c6ef909150cf91473f27d6ee2e9a3. Change-Id: I707a3c215f37376086e22eaa43916afeed6da4c7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/388456 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
2021-03-23 18:16:52 +00:00
OpDecorate %232 RelaxedPrecision
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
OpDecorate %233 RelaxedPrecision
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
%float = OpTypeFloat 32
%v4float = OpTypeVector %float 4
%_ptr_Output_v4float = OpTypePointer Output %v4float
%sk_FragColor = OpVariable %_ptr_Output_v4float Output
%bool = OpTypeBool
%_ptr_Input_bool = OpTypePointer Input %bool
%sk_Clockwise = OpVariable %_ptr_Input_bool Input
Change sksl tests to avoid SPIR-V validation errors 'in' variables without locations aren't allowed. Use uniforms instead. Bug: skia:11738 Change-Id: Ic066106deb7409cff154b4be7cfb3e03a7025c7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385000 Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: John Stiles <johnstiles@google.com>
2021-03-15 18:58:57 +00:00
%_UniformBuffer = OpTypeStruct %v4float %v4float
%_ptr_Uniform__UniformBuffer = OpTypePointer Uniform %_UniformBuffer
%11 = OpVariable %_ptr_Uniform__UniformBuffer Uniform
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
%v2float = OpTypeVector %float 2
%_ptr_Function_v2float = OpTypePointer Function %v2float
Use op cache for pointer and function types. Previously, we stringized the types and put them into fTypeMap. Using the op cache is a simpler mechanism that should work equally well. Output diffs are almost all ID reorderings. In a few cases we managed to deduplicate function types that stringize differently but come out the same in SPIR-V (e.g. no float/half distinction). Change-Id: If7de5b2dafa12d05c3c2c497a243e9e3908dfee7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/529805 Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Brian Osman <brianosman@google.com>
2022-04-13 18:37:36 +00:00
%16 = OpTypeFunction %float %_ptr_Function_v2float %_ptr_Function_v2float
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
%float_2 = OpConstant %float 2
%float_1 = OpConstant %float 1
%float_4 = OpConstant %float 4
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%_ptr_Function_float = OpTypePointer Function %float
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
%float_3 = OpConstant %float 3
%float_6 = OpConstant %float 6
%float_12 = OpConstant %float 12
%float_16 = OpConstant %float 16
%void = OpTypeVoid
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
%175 = OpTypeFunction %void
Change sksl tests to avoid SPIR-V validation errors 'in' variables without locations aren't allowed. Use uniforms instead. Bug: skia:11738 Change-Id: Ic066106deb7409cff154b4be7cfb3e03a7025c7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385000 Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: John Stiles <johnstiles@google.com>
2021-03-15 18:58:57 +00:00
%_ptr_Uniform_v4float = OpTypePointer Uniform %v4float
%int = OpTypeInt 32 1
%int_1 = OpConstant %int 1
Revert "Eliminate inliner temporary variables for functions with a single exit." This reverts commit 345d72124d35b7c434be7bd9b6b16a66fbd980e6. Reason for revert: can break ES2 for-loop rules Original change's description: > Eliminate inliner temporary variables for functions with a single exit. > > When we determine that a function only contains a single return > statement, there is no need to create a temporary variable and store the > result expression into a variable. Instead, we can directly replace the > function-call expression with the return-statement's expression. > > This dramatically simplifies the final optimized output from chains of > very simple inlined functions, which is a very common pattern for trees > of Skia fragment processors. > > Change-Id: I6789064a321daf43db2e1cef4915f25ed74d6131 > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/344665 > Commit-Queue: John Stiles <johnstiles@google.com> > Reviewed-by: Brian Osman <brianosman@google.com> > Auto-Submit: John Stiles <johnstiles@google.com> TBR=brianosman@google.com,ethannicholas@google.com,johnstiles@google.com Change-Id: I60845f22159605a06047b030e2686a769121a35a No-Presubmit: true No-Tree-Checks: true No-Try: true Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345120 Reviewed-by: John Stiles <johnstiles@google.com> Commit-Queue: John Stiles <johnstiles@google.com>
2020-12-16 18:24:48 +00:00
%float_0 = OpConstant %float 0
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
%_ptr_Function_v4float = OpTypePointer Function %v4float
Change sksl tests to avoid SPIR-V validation errors 'in' variables without locations aren't allowed. Use uniforms instead. Bug: skia:11738 Change-Id: Ic066106deb7409cff154b4be7cfb3e03a7025c7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/385000 Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: John Stiles <johnstiles@google.com>
2021-03-15 18:58:57 +00:00
%int_0 = OpConstant %int 0
Use op cache for pointer and function types. Previously, we stringized the types and put them into fTypeMap. Using the op cache is a simpler mechanism that should work equally well. Output diffs are almost all ID reorderings. In a few cases we managed to deduplicate function types that stringize differently but come out the same in SPIR-V (e.g. no float/half distinction). Change-Id: If7de5b2dafa12d05c3c2c497a243e9e3908dfee7 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/529805 Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Brian Osman <brianosman@google.com>
2022-04-13 18:37:36 +00:00
%soft_light_component_Qhh2h2 = OpFunction %float None %16
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
%17 = OpFunctionParameter %_ptr_Function_v2float
%18 = OpFunctionParameter %_ptr_Function_v2float
%19 = OpLabel
%DSqd = OpVariable %_ptr_Function_float Function
%DCub = OpVariable %_ptr_Function_float Function
%DaSqd = OpVariable %_ptr_Function_float Function
%DaCub = OpVariable %_ptr_Function_float Function
%21 = OpLoad %v2float %17
%22 = OpCompositeExtract %float %21 0
%23 = OpFMul %float %float_2 %22
%24 = OpLoad %v2float %17
%25 = OpCompositeExtract %float %24 1
%26 = OpFOrdLessThanEqual %bool %23 %25
OpSelectionMerge %29 None
OpBranchConditional %26 %27 %28
%27 = OpLabel
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%30 = OpLoad %v2float %18
%31 = OpCompositeExtract %float %30 0
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
%32 = OpLoad %v2float %18
%33 = OpCompositeExtract %float %32 0
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%34 = OpFMul %float %31 %33
%35 = OpLoad %v2float %17
%36 = OpCompositeExtract %float %35 1
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
%37 = OpLoad %v2float %17
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%38 = OpCompositeExtract %float %37 0
%39 = OpFMul %float %float_2 %38
%40 = OpFSub %float %36 %39
%41 = OpFMul %float %34 %40
%42 = OpLoad %v2float %18
%43 = OpCompositeExtract %float %42 1
%44 = OpFDiv %float %41 %43
%46 = OpLoad %v2float %18
%47 = OpCompositeExtract %float %46 1
%48 = OpFSub %float %float_1 %47
%49 = OpLoad %v2float %17
%50 = OpCompositeExtract %float %49 0
%51 = OpFMul %float %48 %50
%52 = OpFAdd %float %44 %51
%53 = OpLoad %v2float %18
%54 = OpCompositeExtract %float %53 0
[sksl] Special-case unary negation on matrix types in MSL/SPIR-V MSL does not support the unary "-" operator on matrix types. Similarly the SPIR-V OpFNegate/OpSNegate operations only work on scalar and vector type. * An expression such as "-<mat>" is now transformed to "-1.0 * <mat>" when generating MSL. * The same expression now generates a component-wise negation in SPIR-V, matching what glslang outputs for GLSL. * A unary "+" is now treated as NOP for MSL, matching the SPIR-V backend. An expression such as "+<expr>" is now evaluated as "<expr>". * The shared/Negation.sksl has been moved to folding/ as much of its contents exercise constant-folding of comparison expressions. * The shared/UnaryPositiveNegative.sksl test has been extended to exercise scalar and matrix types. NOTE: The SPIR-V backend changes have caused a minor re-ordering of SSA IDs generated when writing out a prefix-expression. The affected gold files have been updated. Bug: skia:12627, skia:12992 Change-Id: Iec5cdafc591aed7e49b3b52bda42a02661380bab Reviewed-on: https://skia-review.googlesource.com/c/skia/+/513976 Auto-Submit: Arman Uguray <armansito@google.com> Reviewed-by: John Stiles <johnstiles@google.com> Commit-Queue: Arman Uguray <armansito@google.com>
2022-03-01 19:39:57 +00:00
%55 = OpLoad %v2float %17
%56 = OpCompositeExtract %float %55 1
%57 = OpFNegate %float %56
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%58 = OpLoad %v2float %17
%59 = OpCompositeExtract %float %58 0
%60 = OpFMul %float %float_2 %59
[sksl] Special-case unary negation on matrix types in MSL/SPIR-V MSL does not support the unary "-" operator on matrix types. Similarly the SPIR-V OpFNegate/OpSNegate operations only work on scalar and vector type. * An expression such as "-<mat>" is now transformed to "-1.0 * <mat>" when generating MSL. * The same expression now generates a component-wise negation in SPIR-V, matching what glslang outputs for GLSL. * A unary "+" is now treated as NOP for MSL, matching the SPIR-V backend. An expression such as "+<expr>" is now evaluated as "<expr>". * The shared/Negation.sksl has been moved to folding/ as much of its contents exercise constant-folding of comparison expressions. * The shared/UnaryPositiveNegative.sksl test has been extended to exercise scalar and matrix types. NOTE: The SPIR-V backend changes have caused a minor re-ordering of SSA IDs generated when writing out a prefix-expression. The affected gold files have been updated. Bug: skia:12627, skia:12992 Change-Id: Iec5cdafc591aed7e49b3b52bda42a02661380bab Reviewed-on: https://skia-review.googlesource.com/c/skia/+/513976 Auto-Submit: Arman Uguray <armansito@google.com> Reviewed-by: John Stiles <johnstiles@google.com> Commit-Queue: Arman Uguray <armansito@google.com>
2022-03-01 19:39:57 +00:00
%61 = OpFAdd %float %57 %60
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%62 = OpFAdd %float %61 %float_1
%63 = OpFMul %float %54 %62
%64 = OpFAdd %float %52 %63
OpReturnValue %64
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
%28 = OpLabel
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%66 = OpLoad %v2float %18
%67 = OpCompositeExtract %float %66 0
%68 = OpFMul %float %float_4 %67
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
%69 = OpLoad %v2float %18
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%70 = OpCompositeExtract %float %69 1
%71 = OpFOrdLessThanEqual %bool %68 %70
OpSelectionMerge %74 None
OpBranchConditional %71 %72 %73
%72 = OpLabel
%77 = OpLoad %v2float %18
%78 = OpCompositeExtract %float %77 0
Eliminate inliner temporary variables for top-level-exit functions. When we determine that a function only contains a single return statement and it is at the top level (i.e. not inside any scopes), there is no need to create a temporary variable and store the result expression into a variable. Instead, we can directly replace the function-call expression with the return-statement's expression. Unlike my previous solution, this does not require variable declarations to be rewritten. The no-scopes limitation makes it slightly less effective in theory, but in practice we still get almost all of the benefit. The no-scope limitation bites us on structures like @if (true) { return x; } else { return y; } Which will optimize away the if, but leave the scope: { return x; } However, this is not a big deal; the biggest wins are single-line helper functions like `guarded_divide` and `unpremul` which retain the full benefit. Change-Id: I7fbb725e65db021b9795c04c816819669815578f Reviewed-on: https://skia-review.googlesource.com/c/skia/+/345167 Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
2020-12-17 19:38:56 +00:00
%79 = OpLoad %v2float %18
%80 = OpCompositeExtract %float %79 0
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%81 = OpFMul %float %78 %80
OpStore %DSqd %81
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
%83 = OpLoad %v2float %18
%84 = OpCompositeExtract %float %83 0
%85 = OpFMul %float %81 %84
OpStore %DCub %85
%87 = OpLoad %v2float %18
%88 = OpCompositeExtract %float %87 1
%89 = OpLoad %v2float %18
%90 = OpCompositeExtract %float %89 1
%91 = OpFMul %float %88 %90
OpStore %DaSqd %91
%93 = OpLoad %v2float %18
%94 = OpCompositeExtract %float %93 1
%95 = OpFMul %float %91 %94
OpStore %DaCub %95
%96 = OpLoad %v2float %17
%97 = OpCompositeExtract %float %96 0
%98 = OpLoad %v2float %18
%99 = OpCompositeExtract %float %98 0
%101 = OpLoad %v2float %17
%102 = OpCompositeExtract %float %101 1
%103 = OpFMul %float %float_3 %102
%105 = OpLoad %v2float %17
%106 = OpCompositeExtract %float %105 0
%107 = OpFMul %float %float_6 %106
%108 = OpFSub %float %103 %107
%109 = OpFSub %float %108 %float_1
%110 = OpFMul %float %99 %109
%111 = OpFSub %float %97 %110
%112 = OpFMul %float %91 %111
%114 = OpLoad %v2float %18
%115 = OpCompositeExtract %float %114 1
%116 = OpFMul %float %float_12 %115
%117 = OpFMul %float %116 %81
%118 = OpLoad %v2float %17
%119 = OpCompositeExtract %float %118 1
%120 = OpLoad %v2float %17
%121 = OpCompositeExtract %float %120 0
%122 = OpFMul %float %float_2 %121
%123 = OpFSub %float %119 %122
%124 = OpFMul %float %117 %123
%125 = OpFAdd %float %112 %124
%127 = OpFMul %float %float_16 %85
%128 = OpLoad %v2float %17
%129 = OpCompositeExtract %float %128 1
%130 = OpLoad %v2float %17
%131 = OpCompositeExtract %float %130 0
%132 = OpFMul %float %float_2 %131
%133 = OpFSub %float %129 %132
%134 = OpFMul %float %127 %133
%135 = OpFSub %float %125 %134
%136 = OpLoad %v2float %17
%137 = OpCompositeExtract %float %136 0
%138 = OpFMul %float %95 %137
%139 = OpFSub %float %135 %138
%140 = OpFDiv %float %139 %91
OpReturnValue %140
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%73 = OpLabel
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
%141 = OpLoad %v2float %18
%142 = OpCompositeExtract %float %141 0
%143 = OpLoad %v2float %17
%144 = OpCompositeExtract %float %143 1
%145 = OpLoad %v2float %17
%146 = OpCompositeExtract %float %145 0
%147 = OpFMul %float %float_2 %146
%148 = OpFSub %float %144 %147
%149 = OpFAdd %float %148 %float_1
%150 = OpFMul %float %142 %149
%151 = OpLoad %v2float %17
%152 = OpCompositeExtract %float %151 0
%153 = OpFAdd %float %150 %152
%155 = OpLoad %v2float %18
%156 = OpCompositeExtract %float %155 1
%157 = OpLoad %v2float %18
%158 = OpCompositeExtract %float %157 0
%159 = OpFMul %float %156 %158
%154 = OpExtInst %float %1 Sqrt %159
%160 = OpLoad %v2float %17
%161 = OpCompositeExtract %float %160 1
%162 = OpLoad %v2float %17
%163 = OpCompositeExtract %float %162 0
%164 = OpFMul %float %float_2 %163
%165 = OpFSub %float %161 %164
%166 = OpFMul %float %154 %165
%167 = OpFSub %float %153 %166
%168 = OpLoad %v2float %18
%169 = OpCompositeExtract %float %168 1
%170 = OpLoad %v2float %17
%171 = OpCompositeExtract %float %170 0
%172 = OpFMul %float %169 %171
%173 = OpFSub %float %167 %172
OpReturnValue %173
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%74 = OpLabel
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpBranch %29
%29 = OpLabel
OpUnreachable
OpFunctionEnd
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
%main = OpFunction %void None %175
%176 = OpLabel
%185 = OpVariable %_ptr_Function_v4float Function
%196 = OpVariable %_ptr_Function_v2float Function
%200 = OpVariable %_ptr_Function_v2float Function
%205 = OpVariable %_ptr_Function_v2float Function
%209 = OpVariable %_ptr_Function_v2float Function
%214 = OpVariable %_ptr_Function_v2float Function
%218 = OpVariable %_ptr_Function_v2float Function
%177 = OpAccessChain %_ptr_Uniform_v4float %11 %int_1
%181 = OpLoad %v4float %177
%182 = OpCompositeExtract %float %181 3
%184 = OpFOrdEqual %bool %182 %float_0
OpSelectionMerge %189 None
OpBranchConditional %184 %187 %188
%187 = OpLabel
%190 = OpAccessChain %_ptr_Uniform_v4float %11 %int_0
%192 = OpLoad %v4float %190
OpStore %185 %192
OpBranch %189
%188 = OpLabel
%193 = OpAccessChain %_ptr_Uniform_v4float %11 %int_0
%194 = OpLoad %v4float %193
%195 = OpVectorShuffle %v2float %194 %194 0 3
OpStore %196 %195
%197 = OpAccessChain %_ptr_Uniform_v4float %11 %int_1
%198 = OpLoad %v4float %197
%199 = OpVectorShuffle %v2float %198 %198 0 3
OpStore %200 %199
%201 = OpFunctionCall %float %soft_light_component_Qhh2h2 %196 %200
%202 = OpAccessChain %_ptr_Uniform_v4float %11 %int_0
%203 = OpLoad %v4float %202
%204 = OpVectorShuffle %v2float %203 %203 1 3
OpStore %205 %204
%206 = OpAccessChain %_ptr_Uniform_v4float %11 %int_1
%207 = OpLoad %v4float %206
%208 = OpVectorShuffle %v2float %207 %207 1 3
OpStore %209 %208
%210 = OpFunctionCall %float %soft_light_component_Qhh2h2 %205 %209
%211 = OpAccessChain %_ptr_Uniform_v4float %11 %int_0
%212 = OpLoad %v4float %211
%213 = OpVectorShuffle %v2float %212 %212 2 3
OpStore %214 %213
%215 = OpAccessChain %_ptr_Uniform_v4float %11 %int_1
%216 = OpLoad %v4float %215
%217 = OpVectorShuffle %v2float %216 %216 2 3
OpStore %218 %217
%219 = OpFunctionCall %float %soft_light_component_Qhh2h2 %214 %218
%220 = OpAccessChain %_ptr_Uniform_v4float %11 %int_0
%221 = OpLoad %v4float %220
%222 = OpCompositeExtract %float %221 3
%223 = OpAccessChain %_ptr_Uniform_v4float %11 %int_0
%224 = OpLoad %v4float %223
%225 = OpCompositeExtract %float %224 3
%226 = OpFSub %float %float_1 %225
%227 = OpAccessChain %_ptr_Uniform_v4float %11 %int_1
Reduce unnecessary scratch variables in Inliner. 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>
2021-03-19 13:39:44 +00:00
%228 = OpLoad %v4float %227
%229 = OpCompositeExtract %float %228 3
Eliminate redundant loads following stores. We now have two functions `writeOpLoad` and `writeOpStore` which are in charge of writing SpvOpLoad and SpvOpStore instructions. `writeOpStore` also keeps track of pointer stores in a "store cache." Subsequent loads from that same pointer will be found in the cache and will return the value stored in that pointer instead. Such a cache definitely cannot work in the face of control flow, so we make the following concessions: - `pruneReachableOps` is now `pruneConditionalOps`. Any pointers that are altered inside a potentially-unreachable block are cleared from the cache entirely. - The entire store cache is cleared at all OpLabels within a loop. The cache also cannot work in the presence of swizzled stores, so we make another significant concession: - The entire store cache is cleared whenever we store into a non-memory pointer (e.g., assigning into a swizzled LValue, such as `foo.xz`). Despite these significant limitations, this manages to dramatically shrink many real-world examples. Change-Id: I0981a0cf7b45b064e153e9ada271494c8e00cad5 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/530054 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: John Stiles <johnstiles@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2022-04-18 14:57:14 +00:00
%230 = OpFMul %float %226 %229
%231 = OpFAdd %float %222 %230
%232 = OpCompositeConstruct %v4float %201 %210 %219 %231
OpStore %185 %232
OpBranch %189
%189 = OpLabel
%233 = OpLoad %v4float %185
OpStore %sk_FragColor %233
Enable SPIR-V disassembly output from skslc. We now have SPIR-V golden outputs for `blend` and `shared` tests. This exposes a handful of SPIR-V limitations for us to address. Change-Id: Ie5278889b8a61432403d06231b17765885bee0ac Bug: skia:10694 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/337182 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>
2020-11-20 21:28:50 +00:00
OpReturn
OpFunctionEnd
Reference in New Issue Copy Permalink