1 ; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=SICI -check-prefix=GCN -check-prefix=FUNC %s
2 ; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=SICI -check-prefix=CIVI -check-prefix=GCN -check-prefix=FUNC %s
3 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN -check-prefix=FUNC %s
5 ; FUNC-LABEL: {{^}}lds_atomic_cmpxchg_ret_i32_offset:
6 ; GCN: v_mov_b32_e32 [[VCMP:v[0-9]+]], 7
7 ; SICI: s_load_dword [[PTR:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xb
8 ; SICI: s_load_dword [[SWAP:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xc
9 ; VI: s_load_dword [[PTR:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x2c
10 ; VI: s_load_dword [[SWAP:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x30
11 ; GCN-DAG: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[PTR]]
12 ; GCN-DAG: v_mov_b32_e32 [[VSWAP:v[0-9]+]], [[SWAP]]
13 ; GCN: ds_cmpst_rtn_b32 [[RESULT:v[0-9]+]], [[VPTR]], [[VCMP]], [[VSWAP]] offset:16
15 define void @lds_atomic_cmpxchg_ret_i32_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr, i32 %swap) nounwind {
16 %gep = getelementptr i32 addrspace(3)* %ptr, i32 4
17 %pair = cmpxchg i32 addrspace(3)* %gep, i32 7, i32 %swap seq_cst monotonic
18 %result = extractvalue { i32, i1 } %pair, 0
19 store i32 %result, i32 addrspace(1)* %out, align 4
23 ; FUNC-LABEL: {{^}}lds_atomic_cmpxchg_ret_i64_offset:
24 ; GCN-DAG: v_mov_b32_e32 v[[LOVCMP:[0-9]+]], 7
25 ; GCN-DAG: v_mov_b32_e32 v[[HIVCMP:[0-9]+]], 0
26 ; SICI: s_load_dword [[PTR:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xb
27 ; SICI: s_load_dwordx2 s{{\[}}[[LOSWAP:[0-9]+]]:[[HISWAP:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0xd
28 ; VI: s_load_dword [[PTR:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x2c
29 ; VI: s_load_dwordx2 s{{\[}}[[LOSWAP:[0-9]+]]:[[HISWAP:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x34
30 ; GCN-DAG: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[PTR]]
31 ; GCN-DAG: v_mov_b32_e32 v[[LOSWAPV:[0-9]+]], s[[LOSWAP]]
32 ; GCN-DAG: v_mov_b32_e32 v[[HISWAPV:[0-9]+]], s[[HISWAP]]
33 ; GCN: ds_cmpst_rtn_b64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[VPTR]], v{{\[}}[[LOVCMP]]:[[HIVCMP]]{{\]}}, v{{\[}}[[LOSWAPV]]:[[HISWAPV]]{{\]}} offset:32
34 ; GCN: buffer_store_dwordx2 [[RESULT]],
36 define void @lds_atomic_cmpxchg_ret_i64_offset(i64 addrspace(1)* %out, i64 addrspace(3)* %ptr, i64 %swap) nounwind {
37 %gep = getelementptr i64 addrspace(3)* %ptr, i32 4
38 %pair = cmpxchg i64 addrspace(3)* %gep, i64 7, i64 %swap seq_cst monotonic
39 %result = extractvalue { i64, i1 } %pair, 0
40 store i64 %result, i64 addrspace(1)* %out, align 8
44 ; FUNC-LABEL: {{^}}lds_atomic_cmpxchg_ret_i32_bad_si_offset
45 ; SI: ds_cmpst_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
46 ; CIVI: ds_cmpst_rtn_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:16
48 define void @lds_atomic_cmpxchg_ret_i32_bad_si_offset(i32 addrspace(1)* %out, i32 addrspace(3)* %ptr, i32 %swap, i32 %a, i32 %b) nounwind {
50 %add = add i32 %sub, 4
51 %gep = getelementptr i32 addrspace(3)* %ptr, i32 %add
52 %pair = cmpxchg i32 addrspace(3)* %gep, i32 7, i32 %swap seq_cst monotonic
53 %result = extractvalue { i32, i1 } %pair, 0
54 store i32 %result, i32 addrspace(1)* %out, align 4
58 ; FUNC-LABEL: {{^}}lds_atomic_cmpxchg_noret_i32_offset:
59 ; SICI: s_load_dword [[PTR:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x9
60 ; SICI: s_load_dword [[SWAP:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0xa
61 ; VI: s_load_dword [[PTR:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x24
62 ; VI: s_load_dword [[SWAP:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x28
63 ; GCN-DAG: v_mov_b32_e32 [[VCMP:v[0-9]+]], 7
64 ; GCN-DAG: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[PTR]]
65 ; GCN-DAG: v_mov_b32_e32 [[VSWAP:v[0-9]+]], [[SWAP]]
66 ; GCN: ds_cmpst_b32 [[VPTR]], [[VCMP]], [[VSWAP]] offset:16
68 define void @lds_atomic_cmpxchg_noret_i32_offset(i32 addrspace(3)* %ptr, i32 %swap) nounwind {
69 %gep = getelementptr i32 addrspace(3)* %ptr, i32 4
70 %pair = cmpxchg i32 addrspace(3)* %gep, i32 7, i32 %swap seq_cst monotonic
71 %result = extractvalue { i32, i1 } %pair, 0
75 ; FUNC-LABEL: {{^}}lds_atomic_cmpxchg_noret_i64_offset:
76 ; SICI: s_load_dword [[PTR:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x9
77 ; SICI: s_load_dwordx2 s{{\[}}[[LOSWAP:[0-9]+]]:[[HISWAP:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0xb
78 ; VI: s_load_dword [[PTR:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x24
79 ; VI: s_load_dwordx2 s{{\[}}[[LOSWAP:[0-9]+]]:[[HISWAP:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x2c
80 ; GCN-DAG: v_mov_b32_e32 v[[LOVCMP:[0-9]+]], 7
81 ; GCN-DAG: v_mov_b32_e32 v[[HIVCMP:[0-9]+]], 0
82 ; GCN-DAG: v_mov_b32_e32 [[VPTR:v[0-9]+]], [[PTR]]
83 ; GCN-DAG: v_mov_b32_e32 v[[LOSWAPV:[0-9]+]], s[[LOSWAP]]
84 ; GCN-DAG: v_mov_b32_e32 v[[HISWAPV:[0-9]+]], s[[HISWAP]]
85 ; GCN: ds_cmpst_b64 [[VPTR]], v{{\[}}[[LOVCMP]]:[[HIVCMP]]{{\]}}, v{{\[}}[[LOSWAPV]]:[[HISWAPV]]{{\]}} offset:32
87 define void @lds_atomic_cmpxchg_noret_i64_offset(i64 addrspace(3)* %ptr, i64 %swap) nounwind {
88 %gep = getelementptr i64 addrspace(3)* %ptr, i32 4
89 %pair = cmpxchg i64 addrspace(3)* %gep, i64 7, i64 %swap seq_cst monotonic
90 %result = extractvalue { i64, i1 } %pair, 0