AuroraMiddleware/skia2
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
248f57b39b
skia2/tests/sksl/shared/Assignment.asm.frag

222 lines
8.4 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
OpEntryPoint Fragment %main "main" %sk_FragColor %sk_Clockwise
OpExecutionMode %main OriginUpperLeft
OpName %sk_FragColor "sk_FragColor"
OpName %sk_Clockwise "sk_Clockwise"
OpName %main "main"
OpName %x "x"
OpName %ai "ai"
OpName %ai4 "ai4"
OpName %ah2x4 "ah2x4"
OpName %af4 "af4"
OpName %S "S"
OpMemberName %S 0 "f"
OpMemberName %S 1 "af"
OpMemberName %S 2 "h4"
OpMemberName %S 3 "ah4"
OpName %s "s"
OpDecorate %sk_FragColor RelaxedPrecision
OpDecorate %sk_FragColor Location 0
OpDecorate %sk_FragColor Index 0
OpDecorate %sk_Clockwise RelaxedPrecision
OpDecorate %sk_Clockwise BuiltIn FrontFacing
OpDecorate %23 RelaxedPrecision
OpDecorate %_arr_int_int_1 ArrayStride 16
OpDecorate %_arr_v4int_int_1 ArrayStride 16
OpDecorate %_arr_mat2v4float_int_1 ArrayStride 32
OpDecorate %53 RelaxedPrecision
OpDecorate %54 RelaxedPrecision
OpDecorate %52 RelaxedPrecision
OpDecorate %_arr_v4float_int_1 ArrayStride 16
OpDecorate %_arr_float_int_5 ArrayStride 16
OpDecorate %_arr_v4float_int_5 ArrayStride 16
OpMemberDecorate %S 0 Offset 0
OpMemberDecorate %S 1 Offset 16
OpMemberDecorate %S 2 Offset 96
OpMemberDecorate %S 2 RelaxedPrecision
OpMemberDecorate %S 3 Offset 112
OpMemberDecorate %S 3 RelaxedPrecision
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
OpDecorate %83 RelaxedPrecision
OpDecorate %87 RelaxedPrecision
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
OpDecorate %114 RelaxedPrecision
OpDecorate %131 RelaxedPrecision
OpDecorate %141 RelaxedPrecision
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
%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
%void = OpTypeVoid
%11 = OpTypeFunction %void
%_ptr_Function_v4float = OpTypePointer Function %v4float
%float_0 = OpConstant %float 0
%_ptr_Function_float = OpTypePointer Function %float
%int = OpTypeInt 32 1
%int_3 = OpConstant %int 3
%v2float = OpTypeVector %float 2
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%21 = OpConstantComposite %v2float %float_0 %float_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
%int_1 = OpConstant %int 1
%_arr_int_int_1 = OpTypeArray %int %int_1
%_ptr_Function__arr_int_int_1 = OpTypePointer Function %_arr_int_int_1
%int_0 = OpConstant %int 0
%_ptr_Function_int = OpTypePointer Function %int
%v4int = OpTypeVector %int 4
%_arr_v4int_int_1 = OpTypeArray %v4int %int_1
%_ptr_Function__arr_v4int_int_1 = OpTypePointer Function %_arr_v4int_int_1
%int_2 = OpConstant %int 2
%int_4 = OpConstant %int 4
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%37 = OpConstantComposite %v4int %int_1 %int_2 %int_3 %int_4
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
%_ptr_Function_v4int = OpTypePointer Function %v4int
%mat2v4float = OpTypeMatrix %v4float 2
%_arr_mat2v4float_int_1 = OpTypeArray %mat2v4float %int_1
%_ptr_Function__arr_mat2v4float_int_1 = OpTypePointer Function %_arr_mat2v4float_int_1
%float_1 = OpConstant %float 1
%float_2 = OpConstant %float 2
%float_3 = OpConstant %float 3
%float_4 = OpConstant %float 4
%float_5 = OpConstant %float 5
%float_6 = OpConstant %float 6
%float_7 = OpConstant %float 7
%float_8 = OpConstant %float 8
%_ptr_Function_mat2v4float = OpTypePointer Function %mat2v4float
%_arr_v4float_int_1 = OpTypeArray %v4float %int_1
%_ptr_Function__arr_v4float_int_1 = OpTypePointer Function %_arr_v4float_int_1
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%66 = OpConstantComposite %v4float %float_1 %float_1 %float_1 %float_1
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
%int_5 = OpConstant %int 5
%_arr_float_int_5 = OpTypeArray %float %int_5
%_arr_v4float_int_5 = OpTypeArray %v4float %int_5
%S = OpTypeStruct %float %_arr_float_int_5 %v4float %_arr_v4float_int_5
%_ptr_Function_S = OpTypePointer Function %S
%v3float = OpTypeVector %float 3
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%float_9 = OpConstant %float 9
%80 = OpConstantComposite %v3float %float_9 %float_9 %float_9
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%84 = OpConstantComposite %v2float %float_5 %float_5
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%91 = OpConstantComposite %v4float %float_2 %float_2 %float_2 %float_2
%93 = OpConstantComposite %v4float %float_0 %float_0 %float_0 %float_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
%mat3v3float = OpTypeMatrix %v3float 3
%_ptr_Function_mat3v3float = OpTypePointer Function %mat3v3float
%_ptr_Function_v3float = OpTypePointer Function %v3float
%main = OpFunction %void None %11
%12 = OpLabel
%x = OpVariable %_ptr_Function_v4float Function
%ai = OpVariable %_ptr_Function__arr_int_int_1 Function
%ai4 = OpVariable %_ptr_Function__arr_v4int_int_1 Function
%ah2x4 = OpVariable %_ptr_Function__arr_mat2v4float_int_1 Function
%af4 = OpVariable %_ptr_Function__arr_v4float_int_1 Function
%s = OpVariable %_ptr_Function_S Function
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%103 = OpVariable %_ptr_Function_mat3v3float Function
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
%16 = OpAccessChain %_ptr_Function_float %x %int_3
OpStore %16 %float_0
%22 = OpLoad %v4float %x
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%23 = OpVectorShuffle %v4float %22 %21 5 4 2 3
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
OpStore %x %23
%29 = OpAccessChain %_ptr_Function_int %ai %int_0
OpStore %29 %int_0
%38 = OpAccessChain %_ptr_Function_v4int %ai4 %int_0
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
OpStore %38 %37
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
%53 = OpCompositeConstruct %v4float %float_1 %float_2 %float_3 %float_4
%54 = OpCompositeConstruct %v4float %float_5 %float_6 %float_7 %float_8
%52 = OpCompositeConstruct %mat2v4float %53 %54
%55 = OpAccessChain %_ptr_Function_mat2v4float %ah2x4 %int_0
OpStore %55 %52
%57 = OpAccessChain %_ptr_Function_int %ai %int_0
OpStore %57 %int_0
%58 = OpAccessChain %_ptr_Function_int %ai %int_0
%59 = OpLoad %int %58
%60 = OpAccessChain %_ptr_Function_int %ai %59
OpStore %60 %int_0
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%64 = OpAccessChain %_ptr_Function_v4float %af4 %int_0
%65 = OpAccessChain %_ptr_Function_float %64 %int_0
OpStore %65 %float_0
%67 = OpAccessChain %_ptr_Function_v4float %af4 %int_0
%68 = OpLoad %v4float %67
%69 = OpVectorShuffle %v4float %68 %66 6 4 7 5
OpStore %67 %69
%76 = OpAccessChain %_ptr_Function_float %s %int_0
OpStore %76 %float_0
%77 = OpAccessChain %_ptr_Function_float %s %int_1 %int_1
OpStore %77 %float_0
%81 = OpAccessChain %_ptr_Function_v4float %s %int_2
%82 = OpLoad %v4float %81
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%83 = OpVectorShuffle %v4float %82 %80 5 6 4 3
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
OpStore %81 %83
%85 = OpAccessChain %_ptr_Function_v4float %s %int_3 %int_2
%86 = OpLoad %v4float %85
%87 = OpVectorShuffle %v4float %86 %84 0 4 2 5
OpStore %85 %87
%88 = OpAccessChain %_ptr_Function_float %s %int_0
OpStore %88 %float_1
%89 = OpAccessChain %_ptr_Function_float %s %int_1 %int_0
OpStore %89 %float_2
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%90 = OpAccessChain %_ptr_Function_v4float %s %int_2
OpStore %90 %66
%92 = OpAccessChain %_ptr_Function_v4float %s %int_3 %int_0
OpStore %92 %91
OpStore %sk_FragColor %93
%94 = OpCompositeExtract %int %37 0
%95 = OpConvertSToF %float %94
%96 = OpCompositeExtract %int %37 1
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%97 = OpConvertSToF %float %96
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%98 = OpCompositeExtract %int %37 2
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%99 = OpConvertSToF %float %98
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%100 = OpCompositeExtract %int %37 3
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%101 = OpConvertSToF %float %100
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%102 = OpCompositeConstruct %v4float %95 %97 %99 %101
OpStore %sk_FragColor %102
%107 = OpCompositeConstruct %v3float %float_1 %float_2 %float_3
%108 = OpCompositeConstruct %v3float %float_4 %float_5 %float_6
%109 = OpCompositeConstruct %v3float %float_7 %float_8 %float_9
%106 = OpCompositeConstruct %mat3v3float %107 %108 %109
OpStore %103 %106
%110 = OpAccessChain %_ptr_Function_v3float %103 %int_0
%112 = OpLoad %v3float %110
%113 = OpVectorShuffle %v4float %112 %112 0 0 1 2
OpStore %sk_FragColor %113
%114 = OpLoad %v4float %x
OpStore %sk_FragColor %114
Factor out SPIR-V typecasting helper functions. The test diffs look scary, but the only actual change is a minor renumbering of IDs. The actual logic is the same. Change-Id: I5ecc26c8581a4c01834932ff0291deba7d9e4618 Bug: skia:11171 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/353622 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-13 22:15:02 +00:00
%115 = OpAccessChain %_ptr_Function_int %ai %int_0
%116 = OpLoad %int %115
%117 = OpConvertSToF %float %116
%118 = OpCompositeConstruct %v4float %117 %117 %117 %117
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
OpStore %sk_FragColor %118
%119 = OpAccessChain %_ptr_Function_v4int %ai4 %int_0
%120 = OpLoad %v4int %119
%121 = OpCompositeExtract %int %120 0
%122 = OpConvertSToF %float %121
%123 = OpCompositeExtract %int %120 1
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%124 = OpConvertSToF %float %123
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%125 = OpCompositeExtract %int %120 2
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%126 = OpConvertSToF %float %125
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%127 = OpCompositeExtract %int %120 3
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%128 = OpConvertSToF %float %127
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%129 = OpCompositeConstruct %v4float %122 %124 %126 %128
OpStore %sk_FragColor %129
%130 = OpAccessChain %_ptr_Function_v4float %ah2x4 %int_0 %int_0
%131 = OpLoad %v4float %130
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
OpStore %sk_FragColor %131
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%132 = OpAccessChain %_ptr_Function_v4float %af4 %int_0
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%133 = OpLoad %v4float %132
OpStore %sk_FragColor %133
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
OpStore %sk_FragColor %93
%134 = OpAccessChain %_ptr_Function_float %s %int_0
%135 = OpLoad %float %134
%136 = OpCompositeConstruct %v4float %135 %135 %135 %135
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
OpStore %sk_FragColor %136
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%137 = OpAccessChain %_ptr_Function_float %s %int_1 %int_1
Avoid emitting duplicate constant values in SPIR-V. Previously, we had constant-value deduplication, based on the SkSL type of the constant. However, we were still generating redundant constants, because we would emit a separate constant for Float(n) and Half(n), or Int(n) and Short(n), even though we generate the exact same instruction for these constants. We now deduplicate based on the type's number-kind, separating constant literals into three categories: floats, signed ints, and unsigned ints. This better matches our type-handling in getActualType. Change-Id: I5777d4b3d567839b7aa72dc8de76908c18fc387e Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350031 Reviewed-by: Ethan Nicholas <ethannicholas@google.com> Commit-Queue: Ethan Nicholas <ethannicholas@google.com> Auto-Submit: John Stiles <johnstiles@google.com>
2021-01-06 16:57:58 +00:00
%138 = OpLoad %float %137
%139 = OpCompositeConstruct %v4float %138 %138 %138 %138
OpStore %sk_FragColor %139
Deduplicate vector constants in SPIR-V output. Many of our shaders generate the same vector constant dozens of times, e.g. Gaussian blur uses float4(1) repeatedly. This change avoids re-emitting redundant vector constants. Change-Id: I22a71cd8b2783fb997f52d485b49031f64ca6d96 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/350701 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Brian Osman <brianosman@google.com> Reviewed-by: Brian Osman <brianosman@google.com>
2021-01-06 18:33:31 +00:00
%140 = OpAccessChain %_ptr_Function_v4float %s %int_2
%141 = OpLoad %v4float %140
OpStore %sk_FragColor %141
%142 = OpAccessChain %_ptr_Function_v4float %s %int_3 %int_0
%143 = OpLoad %v4float %142
OpStore %sk_FragColor %143
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