xref: /llvm-project/llvm/test/CodeGen/SPIRV/structurizer/cf.for.nested.ll (revision 53326ee0cf45fce3f80e2e98638dd27edb20c516)
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
4;
5; int process() {
6;   int val = 0;
7;
8;   for (int i = 0; i < 10; ++i) {
9;     val = val + i;
10;
11;     for (int j = 0; j < 2; ++j) {
12;       for (int k = 0; k < 2; ++k) {
13;         val = val + k;
14;       }
15;
16;       val = val * 2;
17;
18;     }
19;   }
20;   return val;
21; }
22;
23; [numthreads(1, 1, 1)]
24; void main() {
25;   process();
26; }
27
28; CHECK: %[[#func_11:]] = OpFunction %[[#uint:]] DontInline %[[#]]
29; CHECK:    %[[#bb41:]] = OpLabel
30; CHECK:                  OpBranch %[[#bb42:]]
31; CHECK:     %[[#bb42]] = OpLabel
32; CHECK:                  OpLoopMerge %[[#bb43:]] %[[#bb44:]] None
33; CHECK:                  OpBranchConditional %[[#]] %[[#bb45:]] %[[#bb43]]
34  ; CHECK:     %[[#bb43]] = OpLabel
35  ; CHECK:                  OpReturnValue %[[#]]
36
37  ; CHECK:     %[[#bb45]] = OpLabel
38  ; CHECK:                  OpBranch %[[#bb46:]]
39  ; CHECK:     %[[#bb46]] = OpLabel
40  ; CHECK:                  OpLoopMerge %[[#bb47:]] %[[#bb48:]] None
41  ; CHECK:                  OpBranchConditional %[[#]] %[[#bb49:]] %[[#bb47]]
42    ; CHECK:     %[[#bb47]] = OpLabel
43    ; CHECK:                  OpBranch %[[#bb44]]
44    ; CHECK:     %[[#bb44]] = OpLabel
45    ; CHECK:                  OpBranch %[[#bb42]]
46
47    ; CHECK:     %[[#bb49]] = OpLabel
48    ; CHECK:                  OpBranch %[[#bb50:]]
49    ; CHECK:     %[[#bb50]] = OpLabel
50    ; CHECK:                  OpLoopMerge %[[#bb51:]] %[[#bb52:]] None
51    ; CHECK:                  OpBranchConditional %[[#]] %[[#bb53:]] %[[#bb51]]
52      ; CHECK:     %[[#bb51]] = OpLabel
53      ; CHECK:                  OpBranch %[[#bb48]]
54      ; CHECK:     %[[#bb48]] = OpLabel
55      ; CHECK:                  OpBranch %[[#bb46]]
56
57      ; CHECK:     %[[#bb53]] = OpLabel
58      ; CHECK:                  OpBranch %[[#bb52]]
59      ; CHECK:     %[[#bb52]] = OpLabel
60      ; CHECK:                  OpBranch %[[#bb50]]
61; CHECK:                  OpFunctionEnd
62
63; CHECK: %[[#func_37:]] = OpFunction %[[#void:]] DontInline %[[#]]
64; CHECK:    %[[#bb54:]] = OpLabel
65; CHECK:                  OpReturn
66; CHECK:                  OpFunctionEnd
67
68; CHECK: %[[#func_39:]] = OpFunction %[[#void:]] None %[[#]]
69; CHECK:    %[[#bb55:]] = OpLabel
70; CHECK:                  OpReturn
71; CHECK:                  OpFunctionEnd
72
73
74
75target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-G1"
76target triple = "spirv-unknown-vulkan1.3-compute"
77
78; Function Attrs: convergent noinline norecurse nounwind optnone
79define spir_func noundef i32 @_Z7processv() #0 {
80entry:
81  %0 = call token @llvm.experimental.convergence.entry()
82  %val = alloca i32, align 4
83  %i = alloca i32, align 4
84  %j = alloca i32, align 4
85  %k = alloca i32, align 4
86  store i32 0, ptr %val, align 4
87  store i32 0, ptr %i, align 4
88  br label %for.cond
89
90for.cond:                                         ; preds = %for.inc11, %entry
91  %1 = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token %0) ]
92  %2 = load i32, ptr %i, align 4
93  %cmp = icmp slt i32 %2, 10
94  br i1 %cmp, label %for.body, label %for.end13
95
96for.body:                                         ; preds = %for.cond
97  %3 = load i32, ptr %val, align 4
98  %4 = load i32, ptr %i, align 4
99  %add = add nsw i32 %3, %4
100  store i32 %add, ptr %val, align 4
101  store i32 0, ptr %j, align 4
102  br label %for.cond1
103
104for.cond1:                                        ; preds = %for.inc8, %for.body
105  %5 = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token %1) ]
106  %6 = load i32, ptr %j, align 4
107  %cmp2 = icmp slt i32 %6, 2
108  br i1 %cmp2, label %for.body3, label %for.end10
109
110for.body3:                                        ; preds = %for.cond1
111  store i32 0, ptr %k, align 4
112  br label %for.cond4
113
114for.cond4:                                        ; preds = %for.inc, %for.body3
115  %7 = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token %5) ]
116  %8 = load i32, ptr %k, align 4
117  %cmp5 = icmp slt i32 %8, 2
118  br i1 %cmp5, label %for.body6, label %for.end
119
120for.body6:                                        ; preds = %for.cond4
121  %9 = load i32, ptr %val, align 4
122  %10 = load i32, ptr %k, align 4
123  %add7 = add nsw i32 %9, %10
124  store i32 %add7, ptr %val, align 4
125  br label %for.inc
126
127for.inc:                                          ; preds = %for.body6
128  %11 = load i32, ptr %k, align 4
129  %inc = add nsw i32 %11, 1
130  store i32 %inc, ptr %k, align 4
131  br label %for.cond4
132
133for.end:                                          ; preds = %for.cond4
134  %12 = load i32, ptr %val, align 4
135  %mul = mul nsw i32 %12, 2
136  store i32 %mul, ptr %val, align 4
137  br label %for.inc8
138
139for.inc8:                                         ; preds = %for.end
140  %13 = load i32, ptr %j, align 4
141  %inc9 = add nsw i32 %13, 1
142  store i32 %inc9, ptr %j, align 4
143  br label %for.cond1
144
145for.end10:                                        ; preds = %for.cond1
146  br label %for.inc11
147
148for.inc11:                                        ; preds = %for.end10
149  %14 = load i32, ptr %i, align 4
150  %inc12 = add nsw i32 %14, 1
151  store i32 %inc12, ptr %i, align 4
152  br label %for.cond
153
154for.end13:                                        ; preds = %for.cond
155  %15 = load i32, ptr %val, align 4
156  ret i32 %15
157}
158
159; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none)
160declare token @llvm.experimental.convergence.entry() #1
161
162; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none)
163declare token @llvm.experimental.convergence.loop() #1
164
165; Function Attrs: convergent noinline norecurse nounwind optnone
166define internal spir_func void @main() #0 {
167entry:
168  %0 = call token @llvm.experimental.convergence.entry()
169  %call1 = call spir_func noundef i32 @_Z7processv() #3 [ "convergencectrl"(token %0) ]
170  ret void
171}
172
173; Function Attrs: convergent norecurse
174define void @main.1() #2 {
175entry:
176  call void @main()
177  ret void
178}
179
180attributes #0 = { convergent noinline norecurse nounwind optnone "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
181attributes #1 = { convergent nocallback nofree nosync nounwind willreturn memory(none) }
182attributes #2 = { convergent norecurse "frame-pointer"="all" "hlsl.numthreads"="1,1,1" "hlsl.shader"="compute" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
183attributes #3 = { convergent }
184
185!llvm.module.flags = !{!0, !1, !2}
186
187
188!0 = !{i32 1, !"wchar_size", i32 4}
189!1 = !{i32 4, !"dx.disable_optimizations", i32 1}
190!2 = !{i32 7, !"frame-pointer", i32 2}
191
192
193