1 ; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s 2 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s 3 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s 4 5 ; FUNC-LABEL: {{^}}uint_to_fp_i32_to_f32: 6 ; R600-DAG: UINT_TO_FLT * T{{[0-9]+\.[XYZW]}}, KC0[2].Z 7 8 ; SI: v_cvt_f32_u32_e32 9 ; SI: s_endpgm 10 define void @uint_to_fp_i32_to_f32(float addrspace(1)* %out, i32 %in) { 11 %result = uitofp i32 %in to float 12 store float %result, float addrspace(1)* %out 13 ret void 14 } 15 16 ; FUNC-LABEL: {{^}}uint_to_fp_v2i32_to_v2f32: 17 ; R600-DAG: UINT_TO_FLT * T{{[0-9]+\.[XYZW]}}, KC0[2].W 18 ; R600-DAG: UINT_TO_FLT * T{{[0-9]+\.[XYZW]}}, KC0[3].X 19 20 ; SI: v_cvt_f32_u32_e32 21 ; SI: v_cvt_f32_u32_e32 22 ; SI: s_endpgm 23 define void @uint_to_fp_v2i32_to_v2f32(<2 x float> addrspace(1)* %out, <2 x i32> %in) { 24 %result = uitofp <2 x i32> %in to <2 x float> 25 store <2 x float> %result, <2 x float> addrspace(1)* %out 26 ret void 27 } 28 29 ; FUNC-LABEL: {{^}}uint_to_fp_v4i32_to_v4f32: 30 ; R600: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} 31 ; R600: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} 32 ; R600: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} 33 ; R600: UINT_TO_FLT * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}} 34 35 ; SI: v_cvt_f32_u32_e32 36 ; SI: v_cvt_f32_u32_e32 37 ; SI: v_cvt_f32_u32_e32 38 ; SI: v_cvt_f32_u32_e32 39 ; SI: s_endpgm 40 define void @uint_to_fp_v4i32_to_v4f32(<4 x float> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) { 41 %value = load <4 x i32>, <4 x i32> addrspace(1) * %in 42 %result = uitofp <4 x i32> %value to <4 x float> 43 store <4 x float> %result, <4 x float> addrspace(1)* %out 44 ret void 45 } 46 47 ; FUNC-LABEL: {{^}}uint_to_fp_i64_to_f32: 48 ; R600: UINT_TO_FLT 49 ; R600: UINT_TO_FLT 50 ; R600: MULADD_IEEE 51 ; SI: v_cvt_f32_u32_e32 52 ; SI: v_cvt_f32_u32_e32 53 ; SI: v_madmk_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}, 0x4f800000 54 ; SI: s_endpgm 55 define void @uint_to_fp_i64_to_f32(float addrspace(1)* %out, i64 %in) { 56 entry: 57 %0 = uitofp i64 %in to float 58 store float %0, float addrspace(1)* %out 59 ret void 60 } 61 62 ; FUNC-LABEL: {{^}}uint_to_fp_i1_to_f32: 63 ; SI: v_cmp_eq_i32_e64 [[CMP:s\[[0-9]+:[0-9]\]]], 64 ; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, 1.0, [[CMP]] 65 ; SI: buffer_store_dword [[RESULT]], 66 ; SI: s_endpgm 67 define void @uint_to_fp_i1_to_f32(float addrspace(1)* %out, i32 %in) { 68 %cmp = icmp eq i32 %in, 0 69 %fp = uitofp i1 %cmp to float 70 store float %fp, float addrspace(1)* %out, align 4 71 ret void 72 } 73 74 ; FUNC-LABEL: {{^}}uint_to_fp_i1_to_f32_load: 75 ; SI: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, 1.0 76 ; SI: buffer_store_dword [[RESULT]], 77 ; SI: s_endpgm 78 define void @uint_to_fp_i1_to_f32_load(float addrspace(1)* %out, i1 %in) { 79 %fp = uitofp i1 %in to float 80 store float %fp, float addrspace(1)* %out, align 4 81 ret void 82 } 83