1 ; RUN: llc -mtriple=spirv-unknown-vulkan-compute -O0 %s -o - | FileCheck %s --match-full-lines
2 ; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv-unknown-vulkan-compute %s -o - -filetype=obj | spirv-val %}
4 target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-G1"
5 target triple = "spirv-unknown-vulkan-compute"
7 define internal spir_func void @main() #0 {
9 ; CHECK-DAG: OpName %[[#idx:]] "idx"
10 ; CHECK-DAG: OpDecorate %[[#builtin:]] BuiltIn SubgroupLocalInvocationId
11 ; CHECK-DAG: %[[#int_ty:]] = OpTypeInt 32 0
12 ; CHECK-DAG: %[[#bool_ty:]] = OpTypeBool
13 ; CHECK-DAG: %[[#int_0:]] = OpConstant %[[#int_ty]] 0
14 ; CHECK-DAG: %[[#int_10:]] = OpConstant %[[#int_ty]] 10
16 ; CHECK: %[[#entry:]] = OpLabel
17 ; CHECK: %[[#idx]] = OpVariable %[[#]] Function
18 ; CHECK: OpStore %[[#idx]] %[[#int_0]] Aligned 4
19 ; CHECK: OpBranch %[[#while_cond:]]
21 %0 = call token @llvm.experimental.convergence.entry()
22 %idx = alloca i32, align 4
23 store i32 0, ptr %idx, align 4
26 ; CHECK: %[[#while_cond]] = OpLabel
27 ; CHECK: %[[#tmp:]] = OpLoad %[[#int_ty]] %[[#idx]] Aligned 4
28 ; CHECK: %[[#cmp:]] = OpINotEqual %[[#bool_ty]] %[[#tmp]] %[[#int_10]]
29 ; CHECK: OpLoopMerge %[[#new_end:]] %[[#if_end:]] None
30 ; CHECK: OpBranchConditional %[[#cmp]] %[[#while_body:]] %[[#new_end]]
32 %1 = call token @llvm.experimental.convergence.loop() [ "convergencectrl"(token %0) ]
33 %2 = load i32, ptr %idx, align 4
34 %cmp = icmp ne i32 %2, 10
35 br i1 %cmp, label %while.body, label %while.end
37 ; CHECK: %[[#while_body]] = OpLabel
38 ; CHECK-NEXT: %[[#tmp:]] = OpLoad %[[#int_ty]] %[[#builtin]] Aligned 1
39 ; CHECK-NEXT: OpStore %[[#idx]] %[[#tmp]] Aligned 4
40 ; CHECK-NEXT: %[[#tmp:]] = OpLoad %[[#int_ty]] %[[#idx]] Aligned 4
41 ; CHECK-NEXT: %[[#cmp1:]] = OpIEqual %[[#bool_ty]] %[[#tmp]] %[[#int_0]]
42 ; CHECK: OpBranchConditional %[[#cmp1]] %[[#if_then:]] %[[#if_end]]
44 %3 = call i32 @__hlsl_wave_get_lane_index() [ "convergencectrl"(token %1) ]
45 store i32 %3, ptr %idx, align 4
46 %4 = load i32, ptr %idx, align 4
47 %cmp1 = icmp eq i32 %4, 0
48 br i1 %cmp1, label %if.then, label %if.end
50 ; CHECK: %[[#if_end]] = OpLabel
51 ; CHECK: OpBranch %[[#while_cond]]
53 ; CHECK: %[[#if_then]] = OpLabel
54 ; CHECK: %[[#tmp:]] = OpLoad %[[#int_ty]] %[[#builtin]] Aligned 1
55 ; CHECK: OpStore %[[#idx]] %[[#tmp]] Aligned 4
56 ; CHECK: OpBranch %[[#new_end]]
58 %5 = call i32 @__hlsl_wave_get_lane_index() [ "convergencectrl"(token %1) ]
59 store i32 %5, ptr %idx, align 4
65 ; CHECK: %[[#new_end]] = OpLabel
66 ; CHECK: OpBranch %[[#while_end:]]
68 ; CHECK: %[[#while_end]] = OpLabel
77 declare token @llvm.experimental.convergence.entry() #2
78 declare token @llvm.experimental.convergence.loop() #2
79 declare i32 @__hlsl_wave_get_lane_index() #3
81 attributes #0 = { convergent noinline norecurse nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
82 attributes #1 = { convergent norecurse "hlsl.numthreads"="1,1,1" "hlsl.shader"="compute" "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
83 attributes #2 = { convergent nocallback nofree nosync nounwind willreturn memory(none) }
84 attributes #3 = { convergent }
86 !llvm.module.flags = !{!0, !1}
88 !0 = !{i32 1, !"wchar_size", i32 4}
89 !1 = !{i32 4, !"dx.disable_optimizations", i32 1}