1; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s 2; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s 3 4; FIXME: This should be merged with uint_to_fp.ll, but s_uint_to_fp_v2i64 crashes on r600 5 6; FUNC-LABEL: {{^}}s_uint_to_fp_i64_to_f32: 7define void @s_uint_to_fp_i64_to_f32(float addrspace(1)* %out, i64 %in) #0 { 8 %result = uitofp i64 %in to float 9 store float %result, float addrspace(1)* %out 10 ret void 11} 12 13; FUNC-LABEL: {{^}}v_uint_to_fp_i64_to_f32: 14; GCN: {{buffer|flat}}_load_dwordx2 15 16; GCN: v_ffbh_u32 17; GCN: v_ffbh_u32 18; GCN: v_cndmask 19; GCN: v_cndmask 20 21; GCN-DAG: v_cmp_eq_i64 22; GCN-DAG: v_cmp_lt_u64 23 24; GCN: v_add_i32_e32 [[VR:v[0-9]+]] 25; GCN: {{buffer|flat}}_store_dword {{.*}}[[VR]] 26define void @v_uint_to_fp_i64_to_f32(float addrspace(1)* %out, i64 addrspace(1)* %in) #0 { 27 %tid = call i32 @llvm.r600.read.tidig.x() 28 %in.gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid 29 %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid 30 %val = load i64, i64 addrspace(1)* %in.gep 31 %result = uitofp i64 %val to float 32 store float %result, float addrspace(1)* %out.gep 33 ret void 34} 35 36; FUNC-LABEL: {{^}}s_uint_to_fp_v2i64: 37define void @s_uint_to_fp_v2i64(<2 x float> addrspace(1)* %out, <2 x i64> %in) #0{ 38 %result = uitofp <2 x i64> %in to <2 x float> 39 store <2 x float> %result, <2 x float> addrspace(1)* %out 40 ret void 41} 42 43; FUNC-LABEL: {{^}}v_uint_to_fp_v4i64: 44define void @v_uint_to_fp_v4i64(<4 x float> addrspace(1)* %out, <4 x i64> addrspace(1)* %in) #0 { 45 %tid = call i32 @llvm.r600.read.tidig.x() 46 %in.gep = getelementptr <4 x i64>, <4 x i64> addrspace(1)* %in, i32 %tid 47 %out.gep = getelementptr <4 x float>, <4 x float> addrspace(1)* %out, i32 %tid 48 %value = load <4 x i64>, <4 x i64> addrspace(1)* %in.gep 49 %result = uitofp <4 x i64> %value to <4 x float> 50 store <4 x float> %result, <4 x float> addrspace(1)* %out.gep 51 ret void 52} 53 54declare i32 @llvm.r600.read.tidig.x() #1 55 56attributes #0 = { nounwind } 57attributes #1 = { nounwind readnone } 58