|
| 1 | +; RUN: llc -mtriple=spirv-unknown-vulkan-compute -O0 %s -o - | FileCheck %s |
| 2 | +; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv-unknown-vulkan-compute %s -o - -filetype=obj | spirv-val %} |
| 3 | +; RUN: llc -mtriple=spirv-unknown-vulkan-compute -O0 %s -o - | spirv-sim --function=_Z7processv --wave=1 --expects=19 |
| 4 | + |
| 5 | +; |
| 6 | +; int process() { |
| 7 | +; int val = 0; |
| 8 | +; |
| 9 | +; for (int i = 0; i < 10; ++i) { |
| 10 | +; if (i < 5) { |
| 11 | +; continue; |
| 12 | +; } |
| 13 | +; val = i; |
| 14 | +; |
| 15 | +; { |
| 16 | +; continue; |
| 17 | +; } |
| 18 | +; val++; // No SPIR-V should be emitted for this statement. |
| 19 | +; continue; // No SPIR-V should be emitted for this statement. |
| 20 | +; while(true); // No SPIR-V should be emitted for this statement. |
| 21 | +; } |
| 22 | +; |
| 23 | +; ////////////////////////////////////////////////////////////////////////////////////// |
| 24 | +; // Nested for loops with continue statements // |
| 25 | +; // Each continue statement should branch to the corresponding loop's continue block // |
| 26 | +; ////////////////////////////////////////////////////////////////////////////////////// |
| 27 | +; |
| 28 | +; for (int j = 0; j < 10; ++j) { |
| 29 | +; val = j+5; |
| 30 | +; |
| 31 | +; for ( ; val < 20; ++val) { |
| 32 | +; int k = val + j; |
| 33 | +; continue; |
| 34 | +; k++; // No SPIR-V should be emitted for this statement. |
| 35 | +; } |
| 36 | +; |
| 37 | +; val -= 1; |
| 38 | +; continue; |
| 39 | +; continue; // No SPIR-V should be emitted for this statement. |
| 40 | +; val = val*10; // No SPIR-V should be emitted for this statement. |
| 41 | +; } |
| 42 | +; |
| 43 | +; return val; |
| 44 | +; } |
| 45 | +; |
| 46 | +; [numthreads(1, 1, 1)] |
| 47 | +; void main() { |
| 48 | +; process(); |
| 49 | +; } |
| 50 | + |
| 51 | +; CHECK: %[[#func_12:]] = OpFunction %[[#uint:]] DontInline %[[#]] |
| 52 | +; CHECK: %[[#bb44:]] = OpLabel |
| 53 | +; CHECK: OpBranch %[[#bb45:]] |
| 54 | +; CHECK: %[[#bb45:]] = OpLabel |
| 55 | +; CHECK: OpLoopMerge %[[#bb46:]] %[[#bb47:]] None |
| 56 | +; CHECK: OpBranchConditional %[[#]] %[[#bb48:]] %[[#bb46:]] |
| 57 | +; CHECK: %[[#bb46:]] = OpLabel |
| 58 | +; CHECK: OpBranch %[[#bb49:]] |
| 59 | +; CHECK: %[[#bb49:]] = OpLabel |
| 60 | +; CHECK: OpLoopMerge %[[#bb50:]] %[[#bb51:]] None |
| 61 | +; CHECK: OpBranchConditional %[[#]] %[[#bb52:]] %[[#bb50:]] |
| 62 | +; CHECK: %[[#bb50:]] = OpLabel |
| 63 | +; CHECK: OpReturnValue %[[#]] |
| 64 | +; CHECK: %[[#bb52:]] = OpLabel |
| 65 | +; CHECK: OpBranch %[[#bb53:]] |
| 66 | +; CHECK: %[[#bb53:]] = OpLabel |
| 67 | +; CHECK: OpLoopMerge %[[#bb54:]] %[[#bb55:]] None |
| 68 | +; CHECK: OpBranchConditional %[[#]] %[[#bb56:]] %[[#bb54:]] |
| 69 | +; CHECK: %[[#bb54:]] = OpLabel |
| 70 | +; CHECK: OpBranch %[[#bb51:]] |
| 71 | +; CHECK: %[[#bb51:]] = OpLabel |
| 72 | +; CHECK: OpBranch %[[#bb49:]] |
| 73 | +; CHECK: %[[#bb56:]] = OpLabel |
| 74 | +; CHECK: OpBranch %[[#bb55:]] |
| 75 | +; CHECK: %[[#bb55:]] = OpLabel |
| 76 | +; CHECK: OpBranch %[[#bb53:]] |
| 77 | +; CHECK: %[[#bb48:]] = OpLabel |
| 78 | +; CHECK: OpSelectionMerge %[[#bb57:]] None |
| 79 | +; CHECK: OpBranchConditional %[[#]] %[[#bb57:]] %[[#bb58:]] |
| 80 | +; CHECK: %[[#bb58:]] = OpLabel |
| 81 | +; CHECK: OpBranch %[[#bb57:]] |
| 82 | +; CHECK: %[[#bb57:]] = OpLabel |
| 83 | +; CHECK: OpBranch %[[#bb47:]] |
| 84 | +; CHECK: %[[#bb47:]] = OpLabel |
| 85 | +; CHECK: OpBranch %[[#bb45:]] |
| 86 | +; CHECK: OpFunctionEnd |
| 87 | +; CHECK: %[[#func_40:]] = OpFunction %[[#void:]] DontInline %[[#]] |
| 88 | +; CHECK: %[[#bb59:]] = OpLabel |
| 89 | +; CHECK: OpReturn |
| 90 | +; CHECK: OpFunctionEnd |
| 91 | +; CHECK: %[[#func_42:]] = OpFunction %[[#void:]] None %[[#]] |
| 92 | +; CHECK: %[[#bb60:]] = OpLabel |
| 93 | +; CHECK: OpReturn |
| 94 | +; CHECK: OpFunctionEnd |
| 95 | + |
| 96 | + |
| 97 | + |
| 98 | +target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-G1" |
| 99 | +target triple = "spirv-unknown-vulkan1.3-compute" |
| 100 | + |
| 101 | +; Function Attrs: convergent noinline norecurse nounwind optnone |
| 102 | +define spir_func noundef i32 @_Z7processv() #0 { |
| 103 | +entry: |
| 104 | + %0 = call token @llvm.experimental.convergence.entry() |
| 105 | + %val = alloca i32, align 4 |
| 106 | + %i = alloca i32, align 4 |
| 107 | + %j = alloca i32, align 4 |
| 108 | + %k = alloca i32, align 4 |
| 109 | + store i32 0, ptr %val, align 4 |
| 110 | + store i32 0, ptr %i, align 4 |
| 111 | + br label %for.cond |
| 112 | + |
| 113 | +for.cond: ; preds = %for.inc, %entry |
| 114 | + %1 = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token %0) ] |
| 115 | + %2 = load i32, ptr %i, align 4 |
| 116 | + %cmp = icmp slt i32 %2, 10 |
| 117 | + br i1 %cmp, label %for.body, label %for.end |
| 118 | + |
| 119 | +for.body: ; preds = %for.cond |
| 120 | + %3 = load i32, ptr %i, align 4 |
| 121 | + %cmp1 = icmp slt i32 %3, 5 |
| 122 | + br i1 %cmp1, label %if.then, label %if.end |
| 123 | + |
| 124 | +if.then: ; preds = %for.body |
| 125 | + br label %for.inc |
| 126 | + |
| 127 | +if.end: ; preds = %for.body |
| 128 | + %4 = load i32, ptr %i, align 4 |
| 129 | + store i32 %4, ptr %val, align 4 |
| 130 | + br label %for.inc |
| 131 | + |
| 132 | +for.inc: ; preds = %if.end, %if.then |
| 133 | + %5 = load i32, ptr %i, align 4 |
| 134 | + %inc = add nsw i32 %5, 1 |
| 135 | + store i32 %inc, ptr %i, align 4 |
| 136 | + br label %for.cond |
| 137 | + |
| 138 | +for.end: ; preds = %for.cond |
| 139 | + store i32 0, ptr %j, align 4 |
| 140 | + br label %for.cond2 |
| 141 | + |
| 142 | +for.cond2: ; preds = %for.inc12, %for.end |
| 143 | + %6 = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token %0) ] |
| 144 | + %7 = load i32, ptr %j, align 4 |
| 145 | + %cmp3 = icmp slt i32 %7, 10 |
| 146 | + br i1 %cmp3, label %for.body4, label %for.end14 |
| 147 | + |
| 148 | +for.body4: ; preds = %for.cond2 |
| 149 | + %8 = load i32, ptr %j, align 4 |
| 150 | + %add = add nsw i32 %8, 5 |
| 151 | + store i32 %add, ptr %val, align 4 |
| 152 | + br label %for.cond5 |
| 153 | + |
| 154 | +for.cond5: ; preds = %for.inc9, %for.body4 |
| 155 | + %9 = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token %6) ] |
| 156 | + %10 = load i32, ptr %val, align 4 |
| 157 | + %cmp6 = icmp slt i32 %10, 20 |
| 158 | + br i1 %cmp6, label %for.body7, label %for.end11 |
| 159 | + |
| 160 | +for.body7: ; preds = %for.cond5 |
| 161 | + %11 = load i32, ptr %val, align 4 |
| 162 | + %12 = load i32, ptr %j, align 4 |
| 163 | + %add8 = add nsw i32 %11, %12 |
| 164 | + store i32 %add8, ptr %k, align 4 |
| 165 | + br label %for.inc9 |
| 166 | + |
| 167 | +for.inc9: ; preds = %for.body7 |
| 168 | + %13 = load i32, ptr %val, align 4 |
| 169 | + %inc10 = add nsw i32 %13, 1 |
| 170 | + store i32 %inc10, ptr %val, align 4 |
| 171 | + br label %for.cond5 |
| 172 | + |
| 173 | +for.end11: ; preds = %for.cond5 |
| 174 | + %14 = load i32, ptr %val, align 4 |
| 175 | + %sub = sub nsw i32 %14, 1 |
| 176 | + store i32 %sub, ptr %val, align 4 |
| 177 | + br label %for.inc12 |
| 178 | + |
| 179 | +for.inc12: ; preds = %for.end11 |
| 180 | + %15 = load i32, ptr %j, align 4 |
| 181 | + %inc13 = add nsw i32 %15, 1 |
| 182 | + store i32 %inc13, ptr %j, align 4 |
| 183 | + br label %for.cond2 |
| 184 | + |
| 185 | +for.end14: ; preds = %for.cond2 |
| 186 | + %16 = load i32, ptr %val, align 4 |
| 187 | + ret i32 %16 |
| 188 | +} |
| 189 | + |
| 190 | +; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none) |
| 191 | +declare token @llvm.experimental.convergence.entry() #1 |
| 192 | + |
| 193 | +; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none) |
| 194 | +declare token @llvm.experimental.convergence.loop() #1 |
| 195 | + |
| 196 | +; Function Attrs: convergent noinline norecurse nounwind optnone |
| 197 | +define internal spir_func void @main() #0 { |
| 198 | +entry: |
| 199 | + %0 = call token @llvm.experimental.convergence.entry() |
| 200 | + %call1 = call spir_func noundef i32 @_Z7processv() #3 [ "convergencectrl"(token %0) ] |
| 201 | + ret void |
| 202 | +} |
| 203 | + |
| 204 | +; Function Attrs: convergent norecurse |
| 205 | +define void @main.1() #2 { |
| 206 | +entry: |
| 207 | + call void @main() |
| 208 | + ret void |
| 209 | +} |
| 210 | + |
| 211 | +attributes #0 = { convergent noinline norecurse nounwind optnone "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" } |
| 212 | +attributes #1 = { convergent nocallback nofree nosync nounwind willreturn memory(none) } |
| 213 | +attributes #2 = { convergent norecurse "frame-pointer"="all" "hlsl.numthreads"="1,1,1" "hlsl.shader"="compute" "no-trapping-math"="true" "stack-protector-buffer-size"="8" } |
| 214 | +attributes #3 = { convergent } |
| 215 | + |
| 216 | +!llvm.module.flags = !{!0, !1, !2} |
| 217 | + |
| 218 | + |
| 219 | +!0 = !{i32 1, !"wchar_size", i32 4} |
| 220 | +!1 = !{i32 4, !"dx.disable_optimizations", i32 1} |
| 221 | +!2 = !{i32 7, !"frame-pointer", i32 2} |
| 222 | + |
| 223 | + |
0 commit comments