xref: /llvm-project/llvm/test/CodeGen/AMDGPU/array-ptr-calc-i64.ll (revision 9e9907f1cfa424366fba58d9520f9305b537cec9)
1; RUN: llc -mtriple=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
2
3declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32) #0
4declare i32 @llvm.amdgcn.mbcnt.hi(i32, i32) #0
5
6; SI-LABEL: {{^}}test_array_ptr_calc:
7; SI-DAG: v_mul_u32_u24
8; SI-DAG: v_mul_hi_u32_u24
9; SI: s_endpgm
10define amdgpu_kernel void @test_array_ptr_calc(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %inA, ptr addrspace(1) noalias %inB) {
11  %mbcnt.lo = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
12  %tid = call i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %mbcnt.lo)
13  %a_ptr = getelementptr [1025 x i32], ptr addrspace(1) %inA, i32 %tid, i32 0
14  %b_ptr = getelementptr i32, ptr addrspace(1) %inB, i32 %tid
15  %a = load i32, ptr addrspace(1) %a_ptr
16  %b = load i32, ptr addrspace(1) %b_ptr
17  %result = add i32 %a, %b
18  store i32 %result, ptr addrspace(1) %out
19  ret void
20}
21
22attributes #0 = { nounwind readnone }
23