xref: /llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-constant-fold-vector-op.ll (revision 6548b6354d1d990e1c98736f5e7c3de876bedc8e)
1; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2; RUN: llc -O0 -global-isel -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -stop-after=irtranslator -o - %s | FileCheck %s
3
4; The CSEMIRBuilder was mishandling the result operand when constant
5; folding a vector operation, resulting in a missing def for the
6; stored value.
7define amdgpu_kernel void @constant_fold_vector_add() {
8  ; CHECK-LABEL: name: constant_fold_vector_add
9  ; CHECK: bb.1.entry:
10  ; CHECK-NEXT:   liveins: $sgpr8_sgpr9
11  ; CHECK-NEXT: {{  $}}
12  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
13  ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
14  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[C]](s64), [[C]](s64), [[C]](s64), [[C]](s64)
15  ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(p1) = G_CONSTANT i64 0
16  ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
17  ; CHECK-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[C2]](s64), [[C2]](s64), [[C2]](s64), [[C2]](s64)
18  ; CHECK-NEXT:   G_STORE [[BUILD_VECTOR1]](<4 x s64>), [[C1]](p1) :: (store (<4 x s64>) into `ptr addrspace(1) null`, addrspace 1)
19  ; CHECK-NEXT:   S_ENDPGM 0
20entry:
21  %add = add <4 x i64> zeroinitializer, zeroinitializer
22  store <4 x i64> %add, ptr addrspace(1) null, align 32
23  ret void
24}
25
26!llvm.module.flags = !{!0}
27!0 = !{i32 1, !"amdhsa_code_object_version", i32 500}
28