1 ; RUN: llc -O0 -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck %s
3 ; FIXME: Merge into indirect-addressing-si.ll
5 ; Make sure that TwoAddressInstructions keeps src0 as subregister sub0
6 ; of the tied implicit use and def of the super register.
8 ; CHECK-LABEL: {{^}}insert_wo_offset:
9 ; CHECK: s_load_dword [[IN:s[0-9]+]]
10 ; CHECK: s_mov_b32 m0, [[IN]]
11 ; CHECK: v_movreld_b32_e32 v[[ELT0:[0-9]+]]
12 ; CHECK: buffer_store_dwordx4
13 ; CHECK: buffer_store_dwordx4
14 ; CHECK: buffer_store_dwordx4
15 ; CHECK: buffer_store_dwordx4
16 define amdgpu_kernel void @insert_wo_offset(ptr addrspace(1) %out, i32 %in) {
18 %ins = insertelement <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, float 17.0, i32 %in
19 store <16 x float> %ins, ptr addrspace(1) %out
23 ; Make sure we don't hit use of undefined register errors when expanding an
24 ; extract with undef index.
26 ; CHECK-LABEL: {{^}}extract_adjacent_blocks:
27 ; CHECK: s_load_dword [[ARG:s[0-9]+]]
29 ; CHECK: s_cbranch_scc1 [[BB4:.LBB[0-9]+_[0-9]+]]
31 ; CHECK: buffer_load_dwordx4
33 ; CHECK: s_branch [[ENDBB:.LBB[0-9]+_[0-9]+]]
36 ; CHECK: buffer_load_dwordx4
39 ; CHECK: buffer_store_dword
42 define amdgpu_kernel void @extract_adjacent_blocks(i32 %arg) #0 {
44 %tmp = icmp eq i32 %arg, 0
45 br i1 %tmp, label %bb1, label %bb4
48 %tmp2 = load volatile <4 x float>, ptr addrspace(1) undef
49 %tmp3 = extractelement <4 x float> %tmp2, i32 undef
50 call void asm sideeffect "; reg use $0", "v"(<4 x float> %tmp2) #0 ; Prevent block optimize out
54 %tmp5 = load volatile <4 x float>, ptr addrspace(1) undef
55 %tmp6 = extractelement <4 x float> %tmp5, i32 undef
56 call void asm sideeffect "; reg use $0", "v"(<4 x float> %tmp5) #0 ; Prevent block optimize out
60 %tmp8 = phi float [ %tmp3, %bb1 ], [ %tmp6, %bb4 ]
61 store volatile float %tmp8, ptr addrspace(1) undef