1 ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s 2 ; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s 3 4 ; XXX - Why the packing? 5 ; GCN-LABEL: {{^}}scalar_to_vector_v2i32: 6 ; GCN: buffer_load_dword [[VAL:v[0-9]+]], 7 ; GCN: v_lshrrev_b32_e32 [[SHR:v[0-9]+]], 16, [[VAL]] 8 ; GCN: v_lshlrev_b32_e32 [[SHL:v[0-9]+]], 16, [[SHR]] 9 ; GCN: v_or_b32_e32 v[[OR:[0-9]+]], [[SHR]], [[SHL]] 10 ; GCN: v_mov_b32_e32 v[[COPY:[0-9]+]], v[[OR]] 11 ; GCN: buffer_store_dwordx2 v{{\[}}[[OR]]:[[COPY]]{{\]}} 12 define amdgpu_kernel void @scalar_to_vector_v2i32(<4 x i16> addrspace(1)* %out, i32 addrspace(1)* %in) nounwind { 13 %tmp1 = load i32, i32 addrspace(1)* %in, align 4 14 %bc = bitcast i32 %tmp1 to <2 x i16> 15 %tmp2 = shufflevector <2 x i16> %bc, <2 x i16> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1> 16 store <4 x i16> %tmp2, <4 x i16> addrspace(1)* %out, align 8 17 ret void 18 } 19 20 ; GCN-LABEL: {{^}}scalar_to_vector_v2f32: 21 ; GCN: buffer_load_dword [[VAL:v[0-9]+]], 22 ; GCN: v_lshrrev_b32_e32 [[RESULT:v[0-9]+]], 16, [[VAL]] 23 ; GCN: buffer_store_dwordx2 24 define amdgpu_kernel void @scalar_to_vector_v2f32(<4 x i16> addrspace(1)* %out, float addrspace(1)* %in) nounwind { 25 %tmp1 = load float, float addrspace(1)* %in, align 4 26 %bc = bitcast float %tmp1 to <2 x i16> 27 %tmp2 = shufflevector <2 x i16> %bc, <2 x i16> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1> 28 store <4 x i16> %tmp2, <4 x i16> addrspace(1)* %out, align 8 29 ret void 30 } 31 32 ; GCN-LABEL: {{^}}scalar_to_vector_v4i16: 33 ; VI: v_lshlrev_b16_e32 34 ; VI: v_lshlrev_b16_e32 35 ; VI: v_or_b32_e32 36 ; VI: v_lshlrev_b32 37 ; VI: v_or_b32_sdwa 38 ; VI: v_or_b32_sdwa 39 define amdgpu_kernel void @scalar_to_vector_v4i16() { 40 bb: 41 %tmp = load <2 x i8>, <2 x i8> addrspace(1)* undef, align 1 42 %tmp1 = shufflevector <2 x i8> %tmp, <2 x i8> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> 43 %tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <8 x i32> <i32 0, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9> 44 store <8 x i8> %tmp2, <8 x i8> addrspace(1)* undef, align 8 45 ret void 46 } 47 48 ; GCN-LABEL: {{^}}scalar_to_vector_v4f16: 49 ; VI: v_lshlrev_b16_e32 50 ; VI: v_lshlrev_b16_e32 51 ; VI: v_or_b32_e32 52 ; VI: v_lshlrev_b32 53 ; VI: v_or_b32_sdwa 54 ; VI: v_or_b32_sdwa 55 define amdgpu_kernel void @scalar_to_vector_v4f16() { 56 bb: 57 %load = load half, half addrspace(1)* undef, align 1 58 %tmp = bitcast half %load to <2 x i8> 59 %tmp1 = shufflevector <2 x i8> %tmp, <2 x i8> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> 60 %tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <8 x i32> <i32 0, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9> 61 store <8 x i8> %tmp2, <8 x i8> addrspace(1)* undef, align 8 62 ret void 63 } 64 65 ; Getting a SCALAR_TO_VECTOR seems to be tricky. These cases managed 66 ; to produce one, but for some reason never made it to selection. 67 68 69 ; define amdgpu_kernel void @scalar_to_vector_test2(<8 x i8> addrspace(1)* %out, i32 addrspace(1)* %in) nounwind { 70 ; %tmp1 = load i32, i32 addrspace(1)* %in, align 4 71 ; %bc = bitcast i32 %tmp1 to <4 x i8> 72 73 ; %tmp2 = shufflevector <4 x i8> %bc, <4 x i8> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> 74 ; store <8 x i8> %tmp2, <8 x i8> addrspace(1)* %out, align 4 75 ; ret void 76 ; } 77 78 ; define amdgpu_kernel void @scalar_to_vector_test3(<4 x i32> addrspace(1)* %out) nounwind { 79 ; %newvec0 = insertelement <2 x i64> undef, i64 12345, i32 0 80 ; %newvec1 = insertelement <2 x i64> %newvec0, i64 undef, i32 1 81 ; %bc = bitcast <2 x i64> %newvec1 to <4 x i32> 82 ; %add = add <4 x i32> %bc, <i32 1, i32 2, i32 3, i32 4> 83 ; store <4 x i32> %add, <4 x i32> addrspace(1)* %out, align 16 84 ; ret void 85 ; } 86 87 ; define amdgpu_kernel void @scalar_to_vector_test4(<8 x i16> addrspace(1)* %out) nounwind { 88 ; %newvec0 = insertelement <4 x i32> undef, i32 12345, i32 0 89 ; %bc = bitcast <4 x i32> %newvec0 to <8 x i16> 90 ; %add = add <8 x i16> %bc, <i16 1, i16 2, i16 3, i16 4, i16 1, i16 2, i16 3, i16 4> 91 ; store <8 x i16> %add, <8 x i16> addrspace(1)* %out, align 16 92 ; ret void 93 ; } 94 95 ; define amdgpu_kernel void @scalar_to_vector_test5(<4 x i16> addrspace(1)* %out) nounwind { 96 ; %newvec0 = insertelement <2 x i32> undef, i32 12345, i32 0 97 ; %bc = bitcast <2 x i32> %newvec0 to <4 x i16> 98 ; %add = add <4 x i16> %bc, <i16 1, i16 2, i16 3, i16 4> 99 ; store <4 x i16> %add, <4 x i16> addrspace(1)* %out, align 16 100 ; ret void 101 ; } 102 103 define amdgpu_kernel void @scalar_to_vector_test6(<2 x half> addrspace(1)* %out, i8 zeroext %val) nounwind { 104 %newvec0 = insertelement <4 x i8> undef, i8 %val, i32 0 105 %bc = bitcast <4 x i8> %newvec0 to <2 x half> 106 store <2 x half> %bc, <2 x half> addrspace(1)* %out 107 ret void 108 } 109