xref: /llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-temporal-divergent-reg.mir (revision 8a65ee8b2a05e276ab64682c3c9d3f93490cedcc)
1# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 4
2# RUN: llc -global-isel -mtriple=amdgcn-mesa-amdpal -mcpu=gfx1010 -run-pass=amdgpu-global-isel-divergence-lowering -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX10 %s
3
4---
5name: temporal_divergent_i32
6legalized: true
7tracksRegLiveness: true
8body: |
9  ; GFX10-LABEL: name: temporal_divergent_i32
10  ; GFX10: bb.0:
11  ; GFX10-NEXT:   successors: %bb.1(0x80000000)
12  ; GFX10-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2
13  ; GFX10-NEXT: {{  $}}
14  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
15  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
16  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
17  ; GFX10-NEXT:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
18  ; GFX10-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
19  ; GFX10-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 -1
20  ; GFX10-NEXT: {{  $}}
21  ; GFX10-NEXT: bb.1:
22  ; GFX10-NEXT:   successors: %bb.2(0x04000000), %bb.1(0x7c000000)
23  ; GFX10-NEXT: {{  $}}
24  ; GFX10-NEXT:   [[PHI:%[0-9]+]]:_(s32) = G_PHI %7(s32), %bb.1, [[C]](s32), %bb.0
25  ; GFX10-NEXT:   [[PHI1:%[0-9]+]]:_(s32) = G_PHI [[C1]](s32), %bb.0, %9(s32), %bb.1
26  ; GFX10-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
27  ; GFX10-NEXT:   [[ADD:%[0-9]+]]:_(s32) = G_ADD [[PHI1]], [[C2]]
28  ; GFX10-NEXT:   [[UITOFP:%[0-9]+]]:_(s32) = G_UITOFP [[ADD]](s32)
29  ; GFX10-NEXT:   [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[UITOFP]](s32), [[COPY]]
30  ; GFX10-NEXT:   [[INT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[FCMP]](s1), [[PHI]](s32)
31  ; GFX10-NEXT:   SI_LOOP [[INT]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
32  ; GFX10-NEXT:   G_BR %bb.2
33  ; GFX10-NEXT: {{  $}}
34  ; GFX10-NEXT: bb.2:
35  ; GFX10-NEXT:   [[PHI2:%[0-9]+]]:_(s32) = G_PHI [[ADD]](s32), %bb.1
36  ; GFX10-NEXT:   [[PHI3:%[0-9]+]]:_(s32) = G_PHI [[INT]](s32), %bb.1
37  ; GFX10-NEXT:   G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[PHI3]](s32)
38  ; GFX10-NEXT:   G_STORE [[PHI2]](s32), [[MV]](p0) :: (store (s32))
39  ; GFX10-NEXT:   SI_RETURN
40  bb.0:
41    successors: %bb.1(0x80000000)
42    liveins: $vgpr0, $vgpr1, $vgpr2
43
44    %0:_(s32) = COPY $vgpr0
45    %1:_(s32) = COPY $vgpr1
46    %2:_(s32) = COPY $vgpr2
47    %3:_(p0) = G_MERGE_VALUES %1(s32), %2(s32)
48    %4:_(s32) = G_CONSTANT i32 0
49    %5:_(s32) = G_CONSTANT i32 -1
50
51  bb.1:
52    successors: %bb.2(0x04000000), %bb.1(0x7c000000)
53
54    %6:_(s32) = G_PHI %7(s32), %bb.1, %4(s32), %bb.0
55    %8:_(s32) = G_PHI %5(s32), %bb.0, %9(s32), %bb.1
56    %10:_(s32) = G_CONSTANT i32 1
57    %9:_(s32) = G_ADD %8, %10
58    %11:_(s32) = G_UITOFP %9(s32)
59    %12:_(s1) = G_FCMP floatpred(ogt), %11(s32), %0
60    %7:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %12(s1), %6(s32)
61    SI_LOOP %7(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
62    G_BR %bb.2
63
64  bb.2:
65    %13:_(s32) = G_PHI %9(s32), %bb.1
66    %14:_(s32) = G_PHI %7(s32), %bb.1
67    G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %14(s32)
68    G_STORE %13(s32), %3(p0) :: (store (s32))
69    SI_RETURN
70...
71