xref: /llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-amdgpu_vs.ll (revision d9847cde4841140a95404ea7b7d3a57f8bfbf976)
1; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -stop-after=irtranslator %s -o - | FileCheck %s
3
4define amdgpu_vs void @test_f32_inreg(float inreg %arg0) {
5  ; CHECK-LABEL: name: test_f32_inreg
6  ; CHECK: bb.1 (%ir-block.0):
7  ; CHECK-NEXT:   liveins: $sgpr2
8  ; CHECK-NEXT: {{  $}}
9  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
10  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
11  ; CHECK-NEXT:   G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.exp), 32, 15, [[COPY]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), 0, 0
12  ; CHECK-NEXT:   S_ENDPGM 0
13  call void @llvm.amdgcn.exp.f32(i32 32, i32 15, float %arg0, float undef, float undef, float undef, i1 false, i1 false) #0
14  ret void
15}
16
17define amdgpu_vs void @test_f32(float %arg0) {
18  ; CHECK-LABEL: name: test_f32
19  ; CHECK: bb.1 (%ir-block.0):
20  ; CHECK-NEXT:   liveins: $vgpr0
21  ; CHECK-NEXT: {{  $}}
22  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
23  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
24  ; CHECK-NEXT:   G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.exp), 32, 15, [[COPY]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), 0, 0
25  ; CHECK-NEXT:   S_ENDPGM 0
26  call void @llvm.amdgcn.exp.f32(i32 32, i32 15, float %arg0, float undef, float undef, float undef, i1 false, i1 false) #0
27  ret void
28}
29
30define amdgpu_vs void @test_ptr2_inreg(ptr addrspace(4) inreg %arg0) {
31  ; CHECK-LABEL: name: test_ptr2_inreg
32  ; CHECK: bb.1 (%ir-block.0):
33  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3
34  ; CHECK-NEXT: {{  $}}
35  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
36  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
37  ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p4) = G_MERGE_VALUES [[COPY]](s32), [[COPY1]](s32)
38  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[MV]](p4) :: (volatile invariant load (s32) from %ir.arg0, addrspace 4)
39  ; CHECK-NEXT:   S_ENDPGM 0
40  %tmp0 = load volatile i32, ptr addrspace(4) %arg0
41  ret void
42}
43
44define amdgpu_vs void @test_sgpr_alignment0(float inreg %arg0, ptr addrspace(4) inreg %arg1) {
45  ; CHECK-LABEL: name: test_sgpr_alignment0
46  ; CHECK: bb.1 (%ir-block.0):
47  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4
48  ; CHECK-NEXT: {{  $}}
49  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
50  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
51  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
52  ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(p4) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
53  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
54  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[MV]](p4) :: (volatile invariant load (s32) from %ir.arg1, addrspace 4)
55  ; CHECK-NEXT:   G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.exp), 32, 15, [[COPY]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), 0, 0
56  ; CHECK-NEXT:   S_ENDPGM 0
57  %tmp0 = load volatile i32, ptr addrspace(4) %arg1
58  call void @llvm.amdgcn.exp.f32(i32 32, i32 15, float %arg0, float undef, float undef, float undef, i1 false, i1 false) #0
59  ret void
60}
61
62define amdgpu_vs void @test_order(float inreg %arg0, float inreg %arg1, float %arg2, float %arg3) {
63  ; CHECK-LABEL: name: test_order
64  ; CHECK: bb.1 (%ir-block.0):
65  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3, $vgpr0, $vgpr1
66  ; CHECK-NEXT: {{  $}}
67  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
68  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
69  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr0
70  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr1
71  ; CHECK-NEXT:   G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.exp), 32, 15, [[COPY2]](s32), [[COPY]](s32), [[COPY3]](s32), [[COPY1]](s32), 0, 0
72  ; CHECK-NEXT:   S_ENDPGM 0
73  call void @llvm.amdgcn.exp.f32(i32 32, i32 15, float %arg2, float %arg0, float %arg3, float %arg1, i1 false, i1 false) #0
74  ret void
75}
76
77define amdgpu_vs <{ i32, i32 }> @ret_struct(i32 inreg %arg0, i32 inreg %arg1) {
78  ; CHECK-LABEL: name: ret_struct
79  ; CHECK: bb.1.main_body:
80  ; CHECK-NEXT:   liveins: $sgpr2, $sgpr3
81  ; CHECK-NEXT: {{  $}}
82  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
83  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
84  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
85  ; CHECK-NEXT:   [[INTRINSIC_CONVERGENT:%[0-9]+]]:_(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.readfirstlane), [[COPY]](s32)
86  ; CHECK-NEXT:   $sgpr0 = COPY [[INTRINSIC_CONVERGENT]](s32)
87  ; CHECK-NEXT:   [[INTRINSIC_CONVERGENT1:%[0-9]+]]:_(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.readfirstlane), [[COPY1]](s32)
88  ; CHECK-NEXT:   $sgpr1 = COPY [[INTRINSIC_CONVERGENT1]](s32)
89  ; CHECK-NEXT:   SI_RETURN_TO_EPILOG implicit $sgpr0, implicit $sgpr1
90main_body:
91  %tmp0 = insertvalue <{ i32, i32 }> undef, i32 %arg0, 0
92  %tmp1 = insertvalue <{ i32, i32 }> %tmp0, i32 %arg1, 1
93  ret <{ i32, i32 }> %tmp1
94}
95
96define amdgpu_vs i32 @non_void_ret() {
97  ; CHECK-LABEL: name: non_void_ret
98  ; CHECK: bb.1 (%ir-block.0):
99  ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
100  ; CHECK-NEXT:   [[INTRINSIC_CONVERGENT:%[0-9]+]]:_(s32) = G_INTRINSIC_CONVERGENT intrinsic(@llvm.amdgcn.readfirstlane), [[C]](s32)
101  ; CHECK-NEXT:   $sgpr0 = COPY [[INTRINSIC_CONVERGENT]](s32)
102  ; CHECK-NEXT:   SI_RETURN_TO_EPILOG implicit $sgpr0
103  ret i32 0
104}
105
106declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1)  #0
107
108attributes #0 = { nounwind }
109