1; RUN: llc < %s -march=amdgcn -mcpu=tahiti -verify-machineinstrs | FileCheck -check-prefixes=SI,GCN %s 2; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefixes=VI,GCN %s 3 4; GCN-LABEL: {{^}}select0: 5; i64 select should be split into two i32 selects, and we shouldn't need 6; to use a shfit to extract the hi dword of the input. 7; GCN-NOT: s_lshr_b64 8; GCN: v_cndmask 9; GCN: v_cndmask 10define amdgpu_kernel void @select0(i64 addrspace(1)* %out, i32 %cond, i64 %in) { 11entry: 12 %0 = icmp ugt i32 %cond, 5 13 %1 = select i1 %0, i64 0, i64 %in 14 store i64 %1, i64 addrspace(1)* %out 15 ret void 16} 17 18; GCN-LABEL: {{^}}select_trunc_i64: 19; VI: s_cselect_b32 20; VI-NOT: s_cselect_b32 21; SI: v_cndmask_b32 22; SI-NOT: v_cndmask_b32 23define amdgpu_kernel void @select_trunc_i64(i32 addrspace(1)* %out, i32 %cond, i64 %in) nounwind { 24 %cmp = icmp ugt i32 %cond, 5 25 %sel = select i1 %cmp, i64 0, i64 %in 26 %trunc = trunc i64 %sel to i32 27 store i32 %trunc, i32 addrspace(1)* %out, align 4 28 ret void 29} 30 31; GCN-LABEL: {{^}}select_trunc_i64_2: 32; VI: s_cselect_b32 33; VI-NOT: s_cselect_b32 34; SI: v_cndmask_b32 35; SI-NOT: v_cndmask_b32 36define amdgpu_kernel void @select_trunc_i64_2(i32 addrspace(1)* %out, i32 %cond, i64 %a, i64 %b) nounwind { 37 %cmp = icmp ugt i32 %cond, 5 38 %sel = select i1 %cmp, i64 %a, i64 %b 39 %trunc = trunc i64 %sel to i32 40 store i32 %trunc, i32 addrspace(1)* %out, align 4 41 ret void 42} 43 44; GCN-LABEL: {{^}}v_select_trunc_i64_2: 45; VI: s_cselect_b32 46; VI-NOT: s_cselect_b32 47; SI: v_cndmask_b32 48; SI-NOT: v_cndmask_b32 49define amdgpu_kernel void @v_select_trunc_i64_2(i32 addrspace(1)* %out, i32 %cond, i64 addrspace(1)* %aptr, i64 addrspace(1)* %bptr) nounwind { 50 %cmp = icmp ugt i32 %cond, 5 51 %a = load i64, i64 addrspace(1)* %aptr, align 8 52 %b = load i64, i64 addrspace(1)* %bptr, align 8 53 %sel = select i1 %cmp, i64 %a, i64 %b 54 %trunc = trunc i64 %sel to i32 55 store i32 %trunc, i32 addrspace(1)* %out, align 4 56 ret void 57} 58 59; GCN-LABEL: {{^}}v_select_i64_split_imm: 60; GCN-DAG: v_cndmask_b32_e32 {{v[0-9]+}}, 0, {{v[0-9]+}} 61; GCN-DAG: v_cndmask_b32_e32 {{v[0-9]+}}, 63, {{v[0-9]+}} 62; GCN: s_endpgm 63define amdgpu_kernel void @v_select_i64_split_imm(i64 addrspace(1)* %out, i32 %cond, i64 addrspace(1)* %aptr, i64 addrspace(1)* %bptr) nounwind { 64 %cmp = icmp ugt i32 %cond, 5 65 %a = load i64, i64 addrspace(1)* %aptr, align 8 66 %b = load i64, i64 addrspace(1)* %bptr, align 8 67 %sel = select i1 %cmp, i64 %a, i64 270582939648 ; 63 << 32 68 store i64 %sel, i64 addrspace(1)* %out, align 8 69 ret void 70} 71