1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s 3 4define i32 @atomic_nand_i32_lds(i32 addrspace(3)* %ptr) nounwind { 5; GCN-LABEL: atomic_nand_i32_lds: 6; GCN: ; %bb.0: 7; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) 8; GCN-NEXT: ds_read_b32 v1, v0 9; GCN-NEXT: s_mov_b64 s[4:5], 0 10; GCN-NEXT: BB0_1: ; %atomicrmw.start 11; GCN-NEXT: ; =>This Inner Loop Header: Depth=1 12; GCN-NEXT: s_waitcnt lgkmcnt(0) 13; GCN-NEXT: v_mov_b32_e32 v2, v1 14; GCN-NEXT: v_not_b32_e32 v1, v2 15; GCN-NEXT: v_or_b32_e32 v1, -5, v1 16; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) 17; GCN-NEXT: ds_cmpst_rtn_b32 v1, v0, v2, v1 18; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) 19; GCN-NEXT: buffer_wbinvl1_vol 20; GCN-NEXT: v_cmp_eq_u32_e32 vcc, v1, v2 21; GCN-NEXT: s_or_b64 s[4:5], vcc, s[4:5] 22; GCN-NEXT: s_andn2_b64 exec, exec, s[4:5] 23; GCN-NEXT: s_cbranch_execnz BB0_1 24; GCN-NEXT: ; %bb.2: ; %atomicrmw.end 25; GCN-NEXT: s_or_b64 exec, exec, s[4:5] 26; GCN-NEXT: v_mov_b32_e32 v0, v1 27; GCN-NEXT: s_setpc_b64 s[30:31] 28 %result = atomicrmw nand i32 addrspace(3)* %ptr, i32 4 seq_cst 29 ret i32 %result 30} 31 32define i32 @atomic_nand_i32_global(i32 addrspace(1)* %ptr) nounwind { 33; GCN-LABEL: atomic_nand_i32_global: 34; GCN: ; %bb.0: 35; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) 36; GCN-NEXT: global_load_dword v2, v[0:1], off 37; GCN-NEXT: s_mov_b64 s[4:5], 0 38; GCN-NEXT: BB1_1: ; %atomicrmw.start 39; GCN-NEXT: ; =>This Inner Loop Header: Depth=1 40; GCN-NEXT: s_waitcnt vmcnt(0) 41; GCN-NEXT: v_mov_b32_e32 v3, v2 42; GCN-NEXT: v_not_b32_e32 v2, v3 43; GCN-NEXT: v_or_b32_e32 v2, -5, v2 44; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) 45; GCN-NEXT: global_atomic_cmpswap v2, v[0:1], v[2:3], off glc 46; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) 47; GCN-NEXT: buffer_wbinvl1_vol 48; GCN-NEXT: v_cmp_eq_u32_e32 vcc, v2, v3 49; GCN-NEXT: s_or_b64 s[4:5], vcc, s[4:5] 50; GCN-NEXT: s_andn2_b64 exec, exec, s[4:5] 51; GCN-NEXT: s_cbranch_execnz BB1_1 52; GCN-NEXT: ; %bb.2: ; %atomicrmw.end 53; GCN-NEXT: s_or_b64 exec, exec, s[4:5] 54; GCN-NEXT: v_mov_b32_e32 v0, v2 55; GCN-NEXT: s_setpc_b64 s[30:31] 56 %result = atomicrmw nand i32 addrspace(1)* %ptr, i32 4 seq_cst 57 ret i32 %result 58} 59 60define i32 @atomic_nand_i32_flat(i32* %ptr) nounwind { 61; GCN-LABEL: atomic_nand_i32_flat: 62; GCN: ; %bb.0: 63; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) 64; GCN-NEXT: flat_load_dword v2, v[0:1] 65; GCN-NEXT: s_mov_b64 s[4:5], 0 66; GCN-NEXT: BB2_1: ; %atomicrmw.start 67; GCN-NEXT: ; =>This Inner Loop Header: Depth=1 68; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) 69; GCN-NEXT: v_mov_b32_e32 v3, v2 70; GCN-NEXT: v_not_b32_e32 v2, v3 71; GCN-NEXT: v_or_b32_e32 v2, -5, v2 72; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) 73; GCN-NEXT: flat_atomic_cmpswap v2, v[0:1], v[2:3] glc 74; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) 75; GCN-NEXT: buffer_wbinvl1_vol 76; GCN-NEXT: s_waitcnt lgkmcnt(0) 77; GCN-NEXT: v_cmp_eq_u32_e32 vcc, v2, v3 78; GCN-NEXT: s_or_b64 s[4:5], vcc, s[4:5] 79; GCN-NEXT: s_andn2_b64 exec, exec, s[4:5] 80; GCN-NEXT: s_cbranch_execnz BB2_1 81; GCN-NEXT: ; %bb.2: ; %atomicrmw.end 82; GCN-NEXT: s_or_b64 exec, exec, s[4:5] 83; GCN-NEXT: v_mov_b32_e32 v0, v2 84; GCN-NEXT: s_setpc_b64 s[30:31] 85 %result = atomicrmw nand i32* %ptr, i32 4 seq_cst 86 ret i32 %result 87} 88