Add struct index-lookup test.
Somewhat surprisingly, this test is ES2 compliant.
We didn't have any existing tests (AFAICS) which exercised nested
lookups having more than one index expression.
Change-Id: I869edc048744fad3a2a5587c8c9d70fc477c86ae
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/648458
Reviewed-by: Brian Osman <brianosman@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
diff --git a/gn/sksl_tests.gni b/gn/sksl_tests.gni
index 48179e0..1924247 100644
--- a/gn/sksl_tests.gni
+++ b/gn/sksl_tests.gni
@@ -639,6 +639,7 @@
"shared/StaticSwitchWithStaticConditionalBreakInsideBlock.sksl",
"shared/StructArrayFollowedByScalar.sksl",
"shared/StructComparison.sksl",
+ "shared/StructIndexLookup.sksl",
"shared/StructMaxDepth.sksl",
"shared/Structs.sksl",
"shared/StructsInFunctions.sksl",
diff --git a/resources/sksl/BUILD.bazel b/resources/sksl/BUILD.bazel
index f127ced..670b143 100644
--- a/resources/sksl/BUILD.bazel
+++ b/resources/sksl/BUILD.bazel
@@ -939,6 +939,7 @@
"shared/StaticSwitchWithStaticConditionalBreakInsideBlock.sksl",
"shared/StructArrayFollowedByScalar.sksl",
"shared/StructComparison.sksl",
+ "shared/StructIndexLookup.sksl",
"shared/StructMaxDepth.sksl",
"shared/Structs.sksl",
"shared/StructsInFunctions.sksl",
diff --git a/resources/sksl/shared/StructIndexLookup.sksl b/resources/sksl/shared/StructIndexLookup.sksl
new file mode 100644
index 0000000..349b574
--- /dev/null
+++ b/resources/sksl/shared/StructIndexLookup.sksl
@@ -0,0 +1,45 @@
+uniform half4 colorGreen, colorRed;
+
+struct InnerLUT {
+ float3 values;
+};
+struct OuterLUT {
+ InnerLUT inner[3];
+};
+struct Root {
+ OuterLUT outer[3];
+};
+
+half4 main(float2 coords) {
+ Root data;
+ data.outer[0].inner[0].values = float3(1, 10, 100);
+ data.outer[0].inner[1].values = float3(2, 20, 200);
+ data.outer[0].inner[2].values = float3(3, 30, 300);
+ data.outer[1].inner[0].values = float3(4, 40, 400);
+ data.outer[1].inner[1].values = float3(5, 50, 500);
+ data.outer[1].inner[2].values = float3(6, 60, 600);
+ data.outer[2].inner[0].values = float3(7, 70, 700);
+ data.outer[2].inner[1].values = float3(8, 80, 800);
+ data.outer[2].inner[2].values = float3(9, 90, 900);
+
+ float3 expected = float3(0);
+ for (int i=0; i<3; ++i) {
+ for (int j=0; j<3; ++j) {
+ expected += float3(1, 10, 100);
+
+ // Compare the float3 as a group.
+ if (data.outer[i].inner[j].values != expected) {
+ return colorRed;
+ }
+
+ // Compare the float3 as individual units.
+ for (int k=0; k<3; ++k) {
+ if (data.outer[i].inner[j].values[k] != expected[k]) {
+ return colorRed;
+ }
+ }
+ }
+ }
+
+ return colorGreen;
+}
diff --git a/tests/SkSLTest.cpp b/tests/SkSLTest.cpp
index 766168a..f0678bc 100644
--- a/tests/SkSLTest.cpp
+++ b/tests/SkSLTest.cpp
@@ -680,6 +680,7 @@
SKSL_TEST(RP + VM + GPU, kApiLevel_T, StackingVectorCasts, "shared/StackingVectorCasts.sksl")
SKSL_TEST(RP + VM + GPU_ES3, kNever, StaticSwitch, "shared/StaticSwitch.sksl")
SKSL_TEST(RP + VM + GPU, kApiLevel_T, StructArrayFollowedByScalar, "shared/StructArrayFollowedByScalar.sksl")
+SKSL_TEST(RP + VM + GPU, kNextRelease,StructIndexLookup, "shared/StructIndexLookup.sksl")
// TODO(skia:13920): StructComparison currently exposes a bug in SPIR-V codegen.
// SKSL_TEST(RP /* +GPU_ES3 */, kNever, StructComparison, "shared/StructComparison.sksl")
SKSL_TEST(RP + VM + GPU, kApiLevel_T, StructsInFunctions, "shared/StructsInFunctions.sksl")
diff --git a/tests/sksl/shared/StructIndexLookup.asm.frag b/tests/sksl/shared/StructIndexLookup.asm.frag
new file mode 100644
index 0000000..0308175
--- /dev/null
+++ b/tests/sksl/shared/StructIndexLookup.asm.frag
@@ -0,0 +1,237 @@
+OpCapability Shader
+%1 = OpExtInstImport "GLSL.std.450"
+OpMemoryModel Logical GLSL450
+OpEntryPoint Fragment %_entrypoint_v "_entrypoint" %sk_Clockwise %sk_FragColor
+OpExecutionMode %_entrypoint_v OriginUpperLeft
+OpName %sk_Clockwise "sk_Clockwise"
+OpName %sk_FragColor "sk_FragColor"
+OpName %_UniformBuffer "_UniformBuffer"
+OpMemberName %_UniformBuffer 0 "colorGreen"
+OpMemberName %_UniformBuffer 1 "colorRed"
+OpName %_entrypoint_v "_entrypoint_v"
+OpName %main "main"
+OpName %InnerLUT "InnerLUT"
+OpMemberName %InnerLUT 0 "values"
+OpName %OuterLUT "OuterLUT"
+OpMemberName %OuterLUT 0 "inner"
+OpName %Root "Root"
+OpMemberName %Root 0 "outer"
+OpName %data "data"
+OpName %expected "expected"
+OpName %i "i"
+OpName %j "j"
+OpName %k "k"
+OpDecorate %sk_Clockwise BuiltIn FrontFacing
+OpDecorate %sk_FragColor RelaxedPrecision
+OpDecorate %sk_FragColor Location 0
+OpDecorate %sk_FragColor Index 0
+OpMemberDecorate %_UniformBuffer 0 Offset 0
+OpMemberDecorate %_UniformBuffer 0 RelaxedPrecision
+OpMemberDecorate %_UniformBuffer 1 Offset 16
+OpMemberDecorate %_UniformBuffer 1 RelaxedPrecision
+OpDecorate %_UniformBuffer Block
+OpDecorate %10 Binding 0
+OpDecorate %10 DescriptorSet 0
+OpMemberDecorate %InnerLUT 0 Offset 0
+OpDecorate %_arr_InnerLUT_int_3 ArrayStride 16
+OpMemberDecorate %OuterLUT 0 Offset 0
+OpMemberDecorate %OuterLUT 0 RelaxedPrecision
+OpDecorate %_arr_OuterLUT_int_3 ArrayStride 48
+OpMemberDecorate %Root 0 Offset 0
+OpMemberDecorate %Root 0 RelaxedPrecision
+OpDecorate %117 RelaxedPrecision
+OpDecorate %139 RelaxedPrecision
+OpDecorate %147 RelaxedPrecision
+%bool = OpTypeBool
+%_ptr_Input_bool = OpTypePointer Input %bool
+%sk_Clockwise = OpVariable %_ptr_Input_bool Input
+%float = OpTypeFloat 32
+%v4float = OpTypeVector %float 4
+%_ptr_Output_v4float = OpTypePointer Output %v4float
+%sk_FragColor = OpVariable %_ptr_Output_v4float Output
+%_UniformBuffer = OpTypeStruct %v4float %v4float
+%_ptr_Uniform__UniformBuffer = OpTypePointer Uniform %_UniformBuffer
+%10 = OpVariable %_ptr_Uniform__UniformBuffer Uniform
+%void = OpTypeVoid
+%15 = OpTypeFunction %void
+%float_0 = OpConstant %float 0
+%v2float = OpTypeVector %float 2
+%19 = OpConstantComposite %v2float %float_0 %float_0
+%_ptr_Function_v2float = OpTypePointer Function %v2float
+%23 = OpTypeFunction %v4float %_ptr_Function_v2float
+%v3float = OpTypeVector %float 3
+%InnerLUT = OpTypeStruct %v3float
+%int = OpTypeInt 32 1
+%int_3 = OpConstant %int 3
+%_arr_InnerLUT_int_3 = OpTypeArray %InnerLUT %int_3
+%OuterLUT = OpTypeStruct %_arr_InnerLUT_int_3
+%_arr_OuterLUT_int_3 = OpTypeArray %OuterLUT %int_3
+%Root = OpTypeStruct %_arr_OuterLUT_int_3
+%_ptr_Function_Root = OpTypePointer Function %Root
+%float_1 = OpConstant %float 1
+%float_10 = OpConstant %float 10
+%float_100 = OpConstant %float 100
+%39 = OpConstantComposite %v3float %float_1 %float_10 %float_100
+%int_0 = OpConstant %int 0
+%_ptr_Function_v3float = OpTypePointer Function %v3float
+%float_2 = OpConstant %float 2
+%float_20 = OpConstant %float 20
+%float_200 = OpConstant %float 200
+%46 = OpConstantComposite %v3float %float_2 %float_20 %float_200
+%int_1 = OpConstant %int 1
+%float_3 = OpConstant %float 3
+%float_30 = OpConstant %float 30
+%float_300 = OpConstant %float 300
+%52 = OpConstantComposite %v3float %float_3 %float_30 %float_300
+%int_2 = OpConstant %int 2
+%float_4 = OpConstant %float 4
+%float_40 = OpConstant %float 40
+%float_400 = OpConstant %float 400
+%58 = OpConstantComposite %v3float %float_4 %float_40 %float_400
+%float_5 = OpConstant %float 5
+%float_50 = OpConstant %float 50
+%float_500 = OpConstant %float 500
+%63 = OpConstantComposite %v3float %float_5 %float_50 %float_500
+%float_6 = OpConstant %float 6
+%float_60 = OpConstant %float 60
+%float_600 = OpConstant %float 600
+%68 = OpConstantComposite %v3float %float_6 %float_60 %float_600
+%float_7 = OpConstant %float 7
+%float_70 = OpConstant %float 70
+%float_700 = OpConstant %float 700
+%73 = OpConstantComposite %v3float %float_7 %float_70 %float_700
+%float_8 = OpConstant %float 8
+%float_80 = OpConstant %float 80
+%float_800 = OpConstant %float 800
+%78 = OpConstantComposite %v3float %float_8 %float_80 %float_800
+%float_9 = OpConstant %float 9
+%float_90 = OpConstant %float 90
+%float_900 = OpConstant %float 900
+%83 = OpConstantComposite %v3float %float_9 %float_90 %float_900
+%86 = OpConstantComposite %v3float %float_0 %float_0 %float_0
+%_ptr_Function_int = OpTypePointer Function %int
+%v3bool = OpTypeVector %bool 3
+%_ptr_Uniform_v4float = OpTypePointer Uniform %v4float
+%_entrypoint_v = OpFunction %void None %15
+%16 = OpLabel
+%20 = OpVariable %_ptr_Function_v2float Function
+OpStore %20 %19
+%22 = OpFunctionCall %v4float %main %20
+OpStore %sk_FragColor %22
+OpReturn
+OpFunctionEnd
+%main = OpFunction %v4float None %23
+%24 = OpFunctionParameter %_ptr_Function_v2float
+%25 = OpLabel
+%data = OpVariable %_ptr_Function_Root Function
+%expected = OpVariable %_ptr_Function_v3float Function
+%i = OpVariable %_ptr_Function_int Function
+%j = OpVariable %_ptr_Function_int Function
+%k = OpVariable %_ptr_Function_int Function
+%41 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_0 %int_0 %int_0 %int_0
+OpStore %41 %39
+%48 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_0 %int_0 %int_1 %int_0
+OpStore %48 %46
+%54 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_0 %int_0 %int_2 %int_0
+OpStore %54 %52
+%59 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_1 %int_0 %int_0 %int_0
+OpStore %59 %58
+%64 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_1 %int_0 %int_1 %int_0
+OpStore %64 %63
+%69 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_1 %int_0 %int_2 %int_0
+OpStore %69 %68
+%74 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_2 %int_0 %int_0 %int_0
+OpStore %74 %73
+%79 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_2 %int_0 %int_1 %int_0
+OpStore %79 %78
+%84 = OpAccessChain %_ptr_Function_v3float %data %int_0 %int_2 %int_0 %int_2 %int_0
+OpStore %84 %83
+OpStore %expected %86
+OpStore %i %int_0
+OpBranch %89
+%89 = OpLabel
+OpLoopMerge %93 %92 None
+OpBranch %90
+%90 = OpLabel
+%94 = OpLoad %int %i
+%95 = OpSLessThan %bool %94 %int_3
+OpBranchConditional %95 %91 %93
+%91 = OpLabel
+OpStore %j %int_0
+OpBranch %97
+%97 = OpLabel
+OpLoopMerge %101 %100 None
+OpBranch %98
+%98 = OpLabel
+%102 = OpLoad %int %j
+%103 = OpSLessThan %bool %102 %int_3
+OpBranchConditional %103 %99 %101
+%99 = OpLabel
+%104 = OpLoad %v3float %expected
+%105 = OpFAdd %v3float %104 %39
+OpStore %expected %105
+%106 = OpLoad %int %i
+%107 = OpLoad %int %j
+%108 = OpAccessChain %_ptr_Function_v3float %data %int_0 %106 %int_0 %107 %int_0
+%109 = OpLoad %v3float %108
+%110 = OpFUnordNotEqual %v3bool %109 %105
+%112 = OpAny %bool %110
+OpSelectionMerge %114 None
+OpBranchConditional %112 %113 %114
+%113 = OpLabel
+%115 = OpAccessChain %_ptr_Uniform_v4float %10 %int_1
+%117 = OpLoad %v4float %115
+OpReturnValue %117
+%114 = OpLabel
+OpStore %k %int_0
+OpBranch %119
+%119 = OpLabel
+OpLoopMerge %123 %122 None
+OpBranch %120
+%120 = OpLabel
+%124 = OpLoad %int %k
+%125 = OpSLessThan %bool %124 %int_3
+OpBranchConditional %125 %121 %123
+%121 = OpLabel
+%126 = OpLoad %int %i
+%127 = OpLoad %int %j
+%128 = OpAccessChain %_ptr_Function_v3float %data %int_0 %126 %int_0 %127 %int_0
+%129 = OpLoad %v3float %128
+%130 = OpLoad %int %k
+%131 = OpVectorExtractDynamic %float %129 %130
+%132 = OpLoad %v3float %expected
+%133 = OpLoad %int %k
+%134 = OpVectorExtractDynamic %float %132 %133
+%135 = OpFUnordNotEqual %bool %131 %134
+OpSelectionMerge %137 None
+OpBranchConditional %135 %136 %137
+%136 = OpLabel
+%138 = OpAccessChain %_ptr_Uniform_v4float %10 %int_1
+%139 = OpLoad %v4float %138
+OpReturnValue %139
+%137 = OpLabel
+OpBranch %122
+%122 = OpLabel
+%140 = OpLoad %int %k
+%141 = OpIAdd %int %140 %int_1
+OpStore %k %141
+OpBranch %119
+%123 = OpLabel
+OpBranch %100
+%100 = OpLabel
+%142 = OpLoad %int %j
+%143 = OpIAdd %int %142 %int_1
+OpStore %j %143
+OpBranch %97
+%101 = OpLabel
+OpBranch %92
+%92 = OpLabel
+%144 = OpLoad %int %i
+%145 = OpIAdd %int %144 %int_1
+OpStore %i %145
+OpBranch %89
+%93 = OpLabel
+%146 = OpAccessChain %_ptr_Uniform_v4float %10 %int_0
+%147 = OpLoad %v4float %146
+OpReturnValue %147
+OpFunctionEnd
diff --git a/tests/sksl/shared/StructIndexLookup.glsl b/tests/sksl/shared/StructIndexLookup.glsl
new file mode 100644
index 0000000..28a3225
--- /dev/null
+++ b/tests/sksl/shared/StructIndexLookup.glsl
@@ -0,0 +1,40 @@
+
+out vec4 sk_FragColor;
+uniform vec4 colorGreen;
+uniform vec4 colorRed;
+struct InnerLUT {
+ vec3 values;
+};
+struct OuterLUT {
+ InnerLUT inner[3];
+};
+struct Root {
+ OuterLUT outer[3];
+};
+vec4 main() {
+ Root data;
+ data.outer[0].inner[0].values = vec3(1.0, 10.0, 100.0);
+ data.outer[0].inner[1].values = vec3(2.0, 20.0, 200.0);
+ data.outer[0].inner[2].values = vec3(3.0, 30.0, 300.0);
+ data.outer[1].inner[0].values = vec3(4.0, 40.0, 400.0);
+ data.outer[1].inner[1].values = vec3(5.0, 50.0, 500.0);
+ data.outer[1].inner[2].values = vec3(6.0, 60.0, 600.0);
+ data.outer[2].inner[0].values = vec3(7.0, 70.0, 700.0);
+ data.outer[2].inner[1].values = vec3(8.0, 80.0, 800.0);
+ data.outer[2].inner[2].values = vec3(9.0, 90.0, 900.0);
+ vec3 expected = vec3(0.0);
+ for (int i = 0;i < 3; ++i) {
+ for (int j = 0;j < 3; ++j) {
+ expected += vec3(1.0, 10.0, 100.0);
+ if (data.outer[i].inner[j].values != expected) {
+ return colorRed;
+ }
+ for (int k = 0;k < 3; ++k) {
+ if (data.outer[i].inner[j].values[k] != expected[k]) {
+ return colorRed;
+ }
+ }
+ }
+ }
+ return colorGreen;
+}
diff --git a/tests/sksl/shared/StructIndexLookup.hlsl b/tests/sksl/shared/StructIndexLookup.hlsl
new file mode 100644
index 0000000..81beb87
--- /dev/null
+++ b/tests/sksl/shared/StructIndexLookup.hlsl
@@ -0,0 +1,78 @@
+struct InnerLUT
+{
+ float3 values;
+};
+
+struct OuterLUT
+{
+ InnerLUT inner[3];
+};
+
+struct Root
+{
+ OuterLUT outer[3];
+};
+
+cbuffer _UniformBuffer : register(b0, space0)
+{
+ float4 _10_colorGreen : packoffset(c0);
+ float4 _10_colorRed : packoffset(c1);
+};
+
+
+static float4 sk_FragColor;
+
+struct SPIRV_Cross_Output
+{
+ float4 sk_FragColor : SV_Target0;
+};
+
+float4 main(float2 _24)
+{
+ Root data = { { { { { 0.0f.xxx }, { 0.0f.xxx }, { 0.0f.xxx } } }, { { { 0.0f.xxx }, { 0.0f.xxx }, { 0.0f.xxx } } }, { { { 0.0f.xxx }, { 0.0f.xxx }, { 0.0f.xxx } } } } };
+ data.outer[0].inner[0].values = float3(1.0f, 10.0f, 100.0f);
+ data.outer[0].inner[1].values = float3(2.0f, 20.0f, 200.0f);
+ data.outer[0].inner[2].values = float3(3.0f, 30.0f, 300.0f);
+ data.outer[1].inner[0].values = float3(4.0f, 40.0f, 400.0f);
+ data.outer[1].inner[1].values = float3(5.0f, 50.0f, 500.0f);
+ data.outer[1].inner[2].values = float3(6.0f, 60.0f, 600.0f);
+ data.outer[2].inner[0].values = float3(7.0f, 70.0f, 700.0f);
+ data.outer[2].inner[1].values = float3(8.0f, 80.0f, 800.0f);
+ data.outer[2].inner[2].values = float3(9.0f, 90.0f, 900.0f);
+ float3 expected = 0.0f.xxx;
+ for (int i = 0; i < 3; i++)
+ {
+ for (int j = 0; j < 3; j++)
+ {
+ float3 _104 = expected;
+ float3 _105 = _104 + float3(1.0f, 10.0f, 100.0f);
+ expected = _105;
+ if (any(bool3(data.outer[i].inner[j].values.x != _105.x, data.outer[i].inner[j].values.y != _105.y, data.outer[i].inner[j].values.z != _105.z)))
+ {
+ return _10_colorRed;
+ }
+ for (int k = 0; k < 3; k++)
+ {
+ if (data.outer[i].inner[j].values[k] != expected[k])
+ {
+ return _10_colorRed;
+ }
+ }
+ }
+ }
+ return _10_colorGreen;
+}
+
+void frag_main()
+{
+ float2 _20 = 0.0f.xx;
+ sk_FragColor = main(_20);
+}
+
+SPIRV_Cross_Output main()
+{
+ frag_main();
+ SPIRV_Cross_Output stage_output;
+ stage_output.sk_FragColor = sk_FragColor;
+ return stage_output;
+}
diff --git a/tests/sksl/shared/StructIndexLookup.metal b/tests/sksl/shared/StructIndexLookup.metal
new file mode 100644
index 0000000..f16c934
--- /dev/null
+++ b/tests/sksl/shared/StructIndexLookup.metal
@@ -0,0 +1,53 @@
+#include <metal_stdlib>
+#include <simd/simd.h>
+using namespace metal;
+struct InnerLUT {
+ float3 values;
+};
+struct OuterLUT {
+ array<InnerLUT, 3> inner;
+};
+struct Root {
+ array<OuterLUT, 3> outer;
+};
+struct Uniforms {
+ half4 colorGreen;
+ half4 colorRed;
+};
+struct Inputs {
+};
+struct Outputs {
+ half4 sk_FragColor [[color(0)]];
+};
+fragment Outputs fragmentMain(Inputs _in [[stage_in]], constant Uniforms& _uniforms [[buffer(0)]], bool _frontFacing [[front_facing]], float4 _fragCoord [[position]]) {
+ Outputs _out;
+ (void)_out;
+ Root data;
+ data.outer[0].inner[0].values = float3(1.0, 10.0, 100.0);
+ data.outer[0].inner[1].values = float3(2.0, 20.0, 200.0);
+ data.outer[0].inner[2].values = float3(3.0, 30.0, 300.0);
+ data.outer[1].inner[0].values = float3(4.0, 40.0, 400.0);
+ data.outer[1].inner[1].values = float3(5.0, 50.0, 500.0);
+ data.outer[1].inner[2].values = float3(6.0, 60.0, 600.0);
+ data.outer[2].inner[0].values = float3(7.0, 70.0, 700.0);
+ data.outer[2].inner[1].values = float3(8.0, 80.0, 800.0);
+ data.outer[2].inner[2].values = float3(9.0, 90.0, 900.0);
+ float3 expected = float3(0.0);
+ for (int i = 0;i < 3; ++i) {
+ for (int j = 0;j < 3; ++j) {
+ expected += float3(1.0, 10.0, 100.0);
+ if (any(data.outer[i].inner[j].values != expected)) {
+ _out.sk_FragColor = _uniforms.colorRed;
+ return _out;
+ }
+ for (int k = 0;k < 3; ++k) {
+ if (data.outer[i].inner[j].values[k] != expected[k]) {
+ _out.sk_FragColor = _uniforms.colorRed;
+ return _out;
+ }
+ }
+ }
+ }
+ _out.sk_FragColor = _uniforms.colorGreen;
+ return _out;
+}
diff --git a/tests/sksl/shared/StructIndexLookup.skrp b/tests/sksl/shared/StructIndexLookup.skrp
new file mode 100644
index 0000000..7b40cc8
--- /dev/null
+++ b/tests/sksl/shared/StructIndexLookup.skrp
@@ -0,0 +1,143 @@
+ 1. store_src_rg coords = src.rg
+ 2. init_lane_masks CondMask = LoopMask = RetMask = true
+ 3. zero_4_slots_unmasked data.outer[0].inner[0].values, data.outer[0].inner[1].values(0) = 0
+ 4. zero_4_slots_unmasked data.outer[0].inner[1].values(1..2), data.outer[0].inner[2].values(0..1) = 0
+ 5. zero_4_slots_unmasked data.outer[0].inner[2].values(2), data.outer[1].inner[0].values = 0
+ 6. zero_4_slots_unmasked data.outer[1].inner[1].values, data.outer[1].inner[2].values(0) = 0
+ 7. zero_4_slots_unmasked data.outer[1].inner[2].values(1..2), data.outer[2].inner[0].values(0..1) = 0
+ 8. zero_4_slots_unmasked data.outer[2].inner[0].values(2), data.outer[2].inner[1].values = 0
+ 9. zero_3_slots_unmasked data.outer[2].inner[2].values = 0
+ 10. copy_constant $0 = 0x3F800000 (1.0)
+ 11. copy_constant $1 = 0x41200000 (10.0)
+ 12. copy_constant $2 = 0x42C80000 (100.0)
+ 13. copy_3_slots_masked data.outer[0].inner[0].values = Mask($0..2)
+ 14. copy_constant $0 = 0x40000000 (2.0)
+ 15. copy_constant $1 = 0x41A00000 (20.0)
+ 16. copy_constant $2 = 0x43480000 (200.0)
+ 17. copy_3_slots_masked data.outer[0].inner[1].values = Mask($0..2)
+ 18. copy_constant $0 = 0x40400000 (3.0)
+ 19. copy_constant $1 = 0x41F00000 (30.0)
+ 20. copy_constant $2 = 0x43960000 (300.0)
+ 21. copy_3_slots_masked data.outer[0].inner[2].values = Mask($0..2)
+ 22. copy_constant $0 = 0x40800000 (4.0)
+ 23. copy_constant $1 = 0x42200000 (40.0)
+ 24. copy_constant $2 = 0x43C80000 (400.0)
+ 25. copy_3_slots_masked data.outer[1].inner[0].values = Mask($0..2)
+ 26. copy_constant $0 = 0x40A00000 (5.0)
+ 27. copy_constant $1 = 0x42480000 (50.0)
+ 28. copy_constant $2 = 0x43FA0000 (500.0)
+ 29. copy_3_slots_masked data.outer[1].inner[1].values = Mask($0..2)
+ 30. copy_constant $0 = 0x40C00000 (6.0)
+ 31. copy_constant $1 = 0x42700000 (60.0)
+ 32. copy_constant $2 = 0x44160000 (600.0)
+ 33. copy_3_slots_masked data.outer[1].inner[2].values = Mask($0..2)
+ 34. copy_constant $0 = 0x40E00000 (7.0)
+ 35. copy_constant $1 = 0x428C0000 (70.0)
+ 36. copy_constant $2 = 0x442F0000 (700.0)
+ 37. copy_3_slots_masked data.outer[2].inner[0].values = Mask($0..2)
+ 38. copy_constant $0 = 0x41000000 (8.0)
+ 39. copy_constant $1 = 0x42A00000 (80.0)
+ 40. copy_constant $2 = 0x44480000 (800.0)
+ 41. copy_3_slots_masked data.outer[2].inner[1].values = Mask($0..2)
+ 42. copy_constant $0 = 0x41100000 (9.0)
+ 43. copy_constant $1 = 0x42B40000 (90.0)
+ 44. copy_constant $2 = 0x44610000 (900.0)
+ 45. copy_3_slots_masked data.outer[2].inner[2].values = Mask($0..2)
+ 46. zero_4_slots_unmasked expected, i = 0
+ 47. store_loop_mask $0 = LoopMask
+ 48. jump jump +84 (label 0 at #132)
+ 49. label label 0x00000001
+ 50. zero_slot_unmasked j = 0
+ 51. store_loop_mask $1 = LoopMask
+ 52. jump jump +68 (label 2 at #120)
+ 53. label label 0x00000003
+ 54. copy_3_slots_unmasked $2..4 = expected
+ 55. copy_constant $5 = 0x3F800000 (1.0)
+ 56. copy_constant $6 = 0x41200000 (10.0)
+ 57. copy_constant $7 = 0x42C80000 (100.0)
+ 58. add_3_floats $2..4 += $5..7
+ 59. copy_3_slots_masked expected = Mask($2..4)
+ 60. store_condition_mask $2 = CondMask
+ 61. copy_slot_unmasked $13 = i
+ 62. copy_constant $14 = 0x00000009 (1.261169e-44)
+ 63. mul_int $13 *= $14
+ 64. copy_slot_unmasked $11 = j
+ 65. copy_constant $12 = 0x00000003 (4.203895e-45)
+ 66. mul_int $11 *= $12
+ 67. copy_slot_unmasked $12 = $13
+ 68. add_int $11 += $12
+ 69. copy_from_indirect_unmasked $3..5 = Indirect(data.outer[0].inner[0].values + $11)
+ 70. copy_3_slots_unmasked $6..8 = expected
+ 71. cmpne_3_floats $3..5 = notEqual($3..5, $6..8)
+ 72. bitwise_or_int $4 |= $5
+ 73. bitwise_or_int $3 |= $4
+ 74. merge_condition_mask CondMask = $2 & $3
+ 75. copy_4_constants $4..7 = colorRed
+ 76. copy_4_slots_masked [main].result = Mask($4..7)
+ 77. mask_off_return_mask RetMask &= ~(CondMask & LoopMask & RetMask)
+ 78. load_condition_mask CondMask = $2
+ 79. zero_slot_unmasked k = 0
+ 80. store_loop_mask $2 = LoopMask
+ 81. jump jump +27 (label 4 at #108)
+ 82. label label 0x00000005
+ 83. store_condition_mask $3 = CondMask
+ 84. copy_slot_unmasked $13 = i
+ 85. copy_constant $14 = 0x00000009 (1.261169e-44)
+ 86. mul_int $13 *= $14
+ 87. copy_slot_unmasked $11 = j
+ 88. copy_constant $12 = 0x00000003 (4.203895e-45)
+ 89. mul_int $11 *= $12
+ 90. copy_slot_unmasked $12 = $13
+ 91. add_int $11 += $12
+ 92. copy_slot_unmasked $9 = k
+ 93. copy_slot_unmasked $10 = $11
+ 94. add_int $9 += $10
+ 95. copy_from_indirect_unmasked $4 = Indirect(data.outer[0].inner[0].values(0) + $9)
+ 96. copy_slot_unmasked $13 = k
+ 97. copy_from_indirect_unmasked $5 = Indirect(expected(0) + $13)
+ 98. cmpne_float $4 = notEqual($4, $5)
+ 99. merge_condition_mask CondMask = $3 & $4
+ 100. copy_4_constants $5..8 = colorRed
+ 101. copy_4_slots_masked [main].result = Mask($5..8)
+ 102. mask_off_return_mask RetMask &= ~(CondMask & LoopMask & RetMask)
+ 103. load_condition_mask CondMask = $3
+ 104. copy_slot_unmasked $3 = k
+ 105. copy_constant $4 = 0x00000001 (1.401298e-45)
+ 106. add_int $3 += $4
+ 107. copy_slot_masked k = Mask($3)
+ 108. label label 0x00000004
+ 109. copy_slot_unmasked $3 = k
+ 110. copy_constant $4 = 0x00000003 (4.203895e-45)
+ 111. cmplt_int $3 = lessThan($3, $4)
+ 112. merge_loop_mask LoopMask &= $3
+ 113. stack_rewind
+ 114. branch_if_any_active_lanes branch_if_any_active_lanes -32 (label 5 at #82)
+ 115. load_loop_mask LoopMask = $2
+ 116. copy_slot_unmasked $2 = j
+ 117. copy_constant $3 = 0x00000001 (1.401298e-45)
+ 118. add_int $2 += $3
+ 119. copy_slot_masked j = Mask($2)
+ 120. label label 0x00000002
+ 121. copy_slot_unmasked $2 = j
+ 122. copy_constant $3 = 0x00000003 (4.203895e-45)
+ 123. cmplt_int $2 = lessThan($2, $3)
+ 124. merge_loop_mask LoopMask &= $2
+ 125. stack_rewind
+ 126. branch_if_any_active_lanes branch_if_any_active_lanes -73 (label 3 at #53)
+ 127. load_loop_mask LoopMask = $1
+ 128. copy_slot_unmasked $1 = i
+ 129. copy_constant $2 = 0x00000001 (1.401298e-45)
+ 130. add_int $1 += $2
+ 131. copy_slot_masked i = Mask($1)
+ 132. label label 0x00000000
+ 133. copy_slot_unmasked $1 = i
+ 134. copy_constant $2 = 0x00000003 (4.203895e-45)
+ 135. cmplt_int $1 = lessThan($1, $2)
+ 136. merge_loop_mask LoopMask &= $1
+ 137. stack_rewind
+ 138. branch_if_any_active_lanes branch_if_any_active_lanes -89 (label 1 at #49)
+ 139. load_loop_mask LoopMask = $0
+ 140. copy_4_constants $0..3 = colorGreen
+ 141. copy_4_slots_masked [main].result = Mask($0..3)
+ 142. mask_off_return_mask RetMask &= ~(CondMask & LoopMask & RetMask)
+ 143. load_src src.rgba = [main].result
