1//-------------------------------------------------------------------------------------------------- 2// WHEN CREATING A NEW TEST, PLEASE JUST COPY & PASTE WITHOUT EDITS. 3// 4// Set-up that's shared across all tests in this directory. In principle, this 5// config could be moved to lit.local.cfg. However, there are downstream users that 6// do not use these LIT config files. Hence why this is kept inline. 7// 8// DEFINE: %{sparsifier_opts} = enable-runtime-library=true 9// DEFINE: %{sparsifier_opts_sve} = enable-arm-sve=true %{sparsifier_opts} 10// DEFINE: %{compile} = mlir-opt %s --sparsifier="%{sparsifier_opts}" 11// DEFINE: %{compile_sve} = mlir-opt %s --sparsifier="%{sparsifier_opts_sve}" 12// DEFINE: %{run_libs} = -shared-libs=%mlir_c_runner_utils,%mlir_runner_utils 13// DEFINE: %{run_libs_sve} = -shared-libs=%native_mlir_runner_utils,%native_mlir_c_runner_utils 14// DEFINE: %{run_opts} = -e main -entry-point-result=void 15// DEFINE: %{run} = mlir-runner %{run_opts} %{run_libs} 16// DEFINE: %{run_sve} = %mcr_aarch64_cmd --march=aarch64 --mattr="+sve" %{run_opts} %{run_libs_sve} 17// 18// DEFINE: %{env} = 19//-------------------------------------------------------------------------------------------------- 20 21// RUN: %{compile} | %{run} | FileCheck %s 22// 23// Do the same run, but now with direct IR generation. 24// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false enable-buffer-initialization=true 25// RUN: %{compile} | %{run} | FileCheck %s 26// 27// Do the same run, but now with direct IR generation and vectorization. 28// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false enable-buffer-initialization=true vl=2 reassociate-fp-reductions=true enable-index-optimizations=true 29// RUN: %{compile} | %{run} | FileCheck %s 30// 31// Do the same run, but now with direct IR generation and VLA vectorization. 32// RUN: %if mlir_arm_sve_tests %{ %{compile_sve} | %{run_sve} | FileCheck %s %} 33 34#CCC = #sparse_tensor.encoding<{ 35 map = (d0, d1, d2) -> (d0 : compressed, d1 : compressed, d2 : compressed) }> 36 37#CDC = #sparse_tensor.encoding<{ 38 map = (d0, d1, d2) -> (d0 : compressed, d1 : dense, d2 : compressed) 39 40 // FIXME: Still inadmissible might need investigation 41 // dimToLvl = affine_map<(i,j,k) -> (j,k,i)> 42}> 43 44// Creates and returns 3-D buffer of size (%s1, %s2, %s3) filled with the value %f 45func.func @alloc_3d_filled_f32(%s1 : index, %s2 : index, %s3 : index, %f : f32) -> tensor<?x?x?xf32> { 46 %buf = tensor.empty(%s1, %s2, %s3) : tensor<?x?x?xf32> 47 %ret = linalg.fill ins(%f : f32) outs(%buf : tensor<?x?x?xf32>) -> tensor<?x?x?xf32> 48 return %ret : tensor<?x?x?xf32> 49} 50 51func.func @conv_1d_nwc_wcf(%arg0: tensor<?x?x?xf32>, %arg1: tensor<?x?x?xf32>, %arg2: tensor<?x?x?xf32>) -> tensor<?x?x?xf32> { 52 %ret = linalg.conv_1d_nwc_wcf {dilations = dense<1> : tensor<1xi64>, 53 strides = dense<1> : tensor<1xi64>} 54 ins (%arg0, %arg1: tensor<?x?x?xf32>, tensor<?x?x?xf32>) 55 outs (%arg2: tensor<?x?x?xf32>) -> tensor<?x?x?xf32> 56 return %ret : tensor<?x?x?xf32> 57} 58 59func.func @conv_1d_nwc_wcf_CCC(%arg0: tensor<?x?x?xf32, #CCC>, %arg1: tensor<?x?x?xf32>) -> tensor<?x?x?xf32, #CCC> { 60 %c1 = arith.constant 1 : index 61 %c3 = arith.constant 3 : index 62 %c6 = arith.constant 6 : index 63 %s = tensor.empty(%c3, %c6, %c1) : tensor<?x?x?xf32, #CCC> 64 %ret = linalg.conv_1d_nwc_wcf {dilations = dense<1> : tensor<1xi64>, 65 strides = dense<1> : tensor<1xi64>} 66 ins (%arg0, %arg1: tensor<?x?x?xf32, #CCC>, tensor<?x?x?xf32>) 67 outs (%s: tensor<?x?x?xf32, #CCC>) -> tensor<?x?x?xf32, #CCC> 68 return %ret : tensor<?x?x?xf32, #CCC> 69} 70 71func.func @conv_1d_nwc_wcf_CDC(%arg0: tensor<?x?x?xf32, #CDC>, %arg1: tensor<?x?x?xf32>) -> tensor<?x?x?xf32, #CDC> { 72 %c1 = arith.constant 1 : index 73 %c3 = arith.constant 3 : index 74 %c6 = arith.constant 6 : index 75 %s = tensor.empty(%c3, %c6, %c1) : tensor<?x?x?xf32, #CDC> 76 %ret = linalg.conv_1d_nwc_wcf {dilations = dense<1> : tensor<1xi64>, 77 strides = dense<1> : tensor<1xi64>} 78 ins (%arg0, %arg1: tensor<?x?x?xf32, #CDC>, tensor<?x?x?xf32>) 79 outs (%s: tensor<?x?x?xf32, #CDC>) -> tensor<?x?x?xf32, #CDC> 80 return %ret : tensor<?x?x?xf32, #CDC> 81} 82 83func.func @main() { 84 %c0 = arith.constant 0 : index 85 %c1 = arith.constant 1 : index 86 %c3 = arith.constant 3 : index 87 %c6 = arith.constant 6 : index 88 %c8 = arith.constant 8 : index 89 %f10 = arith.constant 10.00000e+00 : f32 90 %val = arith.constant 2.00000e+00 : f32 91 %zero = arith.constant 0.00000e+00 : f32 92 93 %in1D_tmp = call @alloc_3d_filled_f32(%c3, %c8, %c1, %val) : (index, index, index, f32) -> (tensor<?x?x?xf32>) 94 %in1D_nwc = tensor.insert %f10 into %in1D_tmp[%c0, %c3, %c0] : tensor<?x?x?xf32> 95 96 %filter1D_nwc = call @alloc_3d_filled_f32(%c3, %c1, %c1, %val) : (index, index, index, f32) -> (tensor<?x?x?xf32>) 97 %out1D_nwc = call @alloc_3d_filled_f32(%c3, %c6, %c1, %zero) : (index, index, index, f32) -> (tensor<?x?x?xf32>) 98 99 %in1D_nwc_CCC = sparse_tensor.convert %in1D_nwc 100 : tensor<?x?x?xf32> to tensor<?x?x?xf32, #CCC> 101 %in1D_nwc_CDC = sparse_tensor.convert %in1D_nwc 102 : tensor<?x?x?xf32> to tensor<?x?x?xf32, #CDC> 103 104 %dense_ret = call @conv_1d_nwc_wcf(%in1D_nwc, %filter1D_nwc, %out1D_nwc) : (tensor<?x?x?xf32>, tensor<?x?x?xf32>, tensor<?x?x?xf32>) -> (tensor<?x?x?xf32>) 105 %CCC_ret = call @conv_1d_nwc_wcf_CCC(%in1D_nwc_CCC, %filter1D_nwc) : (tensor<?x?x?xf32, #CCC>, tensor<?x?x?xf32>) -> (tensor<?x?x?xf32, #CCC>) 106 %CDC_ret = call @conv_1d_nwc_wcf_CDC(%in1D_nwc_CDC, %filter1D_nwc) : (tensor<?x?x?xf32, #CDC>, tensor<?x?x?xf32>) -> (tensor<?x?x?xf32, #CDC>) 107 108 // CHECK: ( ( ( 12 ), ( 28 ), ( 28 ), ( 28 ), ( 12 ), ( 12 ) ), 109 // CHECK-SAME: ( ( 12 ), ( 12 ), ( 12 ), ( 12 ), ( 12 ), ( 12 ) ), 110 // CHECK-SAME: ( ( 12 ), ( 12 ), ( 12 ), ( 12 ), ( 12 ), ( 12 ) ) ) 111 %dense_v = vector.transfer_read %dense_ret[%c0, %c0, %c0], %zero 112 : tensor<?x?x?xf32>, vector<3x6x1xf32> 113 vector.print %dense_v : vector<3x6x1xf32> 114 115 // 116 // CHECK: ---- Sparse Tensor ---- 117 // CHECK-NEXT: nse = 18 118 // CHECK-NEXT: dim = ( 3, 6, 1 ) 119 // CHECK-NEXT: lvl = ( 3, 6, 1 ) 120 // CHECK-NEXT: pos[0] : ( 0, 3 ) 121 // CHECK-NEXT: crd[0] : ( 0, 1, 2 ) 122 // CHECK-NEXT: pos[1] : ( 0, 6, 12, 18 ) 123 // CHECK-NEXT: crd[1] : ( 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5 ) 124 // CHECK-NEXT: pos[2] : ( 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18 ) 125 // CHECK-NEXT: crd[2] : ( 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ) 126 // CHECK-NEXT: values : ( 12, 28, 28, 28, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12 ) 127 // CHECK-NEXT: ---- 128 // 129 sparse_tensor.print %CCC_ret : tensor<?x?x?xf32, #CCC> 130 131 // 132 // CHECK: ---- Sparse Tensor ---- 133 // CHECK-NEXT: nse = 18 134 // CHECK-NEXT: dim = ( 3, 6, 1 ) 135 // CHECK-NEXT: lvl = ( 3, 6, 1 ) 136 // CHECK-NEXT: pos[0] : ( 0, 3 ) 137 // CHECK-NEXT: crd[0] : ( 0, 1, 2 ) 138 // CHECK-NEXT: pos[2] : ( 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18 ) 139 // CHECK-NEXT: crd[2] : ( 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ) 140 // CHECK-NEXT: values : ( 12, 28, 28, 28, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12 ) 141 // CHECK-NEXT: ---- 142 // 143 sparse_tensor.print %CDC_ret : tensor<?x?x?xf32, #CDC> 144 145 // Free the resources 146 bufferization.dealloc_tensor %in1D_nwc : tensor<?x?x?xf32> 147 bufferization.dealloc_tensor %filter1D_nwc : tensor<?x?x?xf32> 148 bufferization.dealloc_tensor %out1D_nwc : tensor<?x?x?xf32> 149 150 bufferization.dealloc_tensor %in1D_nwc_CDC : tensor<?x?x?xf32, #CDC> 151 bufferization.dealloc_tensor %in1D_nwc_CCC : tensor<?x?x?xf32, #CCC> 152 153 bufferization.dealloc_tensor %CCC_ret : tensor<?x?x?xf32, #CCC> 154 bufferization.dealloc_tensor %CDC_ret : tensor<?x?x?xf32, #CDC> 155 156 return 157} 158