221 lines · plain
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, this5// config could be moved to lit.local.cfg. However, there are downstream users that6// do not use these LIT config files. Hence why this is kept inline.7//8// DEFINE: %{sparsifier_opts} = enable-runtime-library=true9// 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_utils13// DEFINE: %{run_libs_sve} = -shared-libs=%native_mlir_runner_utils,%native_mlir_c_runner_utils14// DEFINE: %{run_opts} = -e main -entry-point-result=void15// 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 %s22//23// Do the same run, but now with direct IR generation.24// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false25// RUN: %{compile} | %{run} | FileCheck %s26//27// Do the same run, but now with vectorization.28// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false vl=2 reassociate-fp-reductions=true enable-index-optimizations=true29// RUN: %{compile} | %{run} | FileCheck %s30//31// Do the same run, but now with VLA vectorization.32// RUN: %if mlir_arm_sve_tests %{ %{compile_sve} | %{run_sve} | FileCheck %s %}33 34//35// Several common sparse storage schemes.36//37 38#Dense = #sparse_tensor.encoding<{39 map = (d0, d1) -> (d0 : dense, d1 : dense)40}>41 42#CSR = #sparse_tensor.encoding<{43 map = (d0, d1) -> (d0 : dense, d1 : compressed)44}>45 46#DCSR = #sparse_tensor.encoding<{47 map = (d0, d1) -> (d0 : compressed, d1 : compressed)48}>49 50#CSC = #sparse_tensor.encoding<{51 map = (d0, d1) -> (d1 : dense, d0 : compressed)52}>53 54#DCSC = #sparse_tensor.encoding<{55 map = (d0, d1) -> (d1 : compressed, d0 : compressed)56}>57 58#BlockRow = #sparse_tensor.encoding<{59 map = (d0, d1) -> (d0 : compressed, d1 : dense)60}>61 62#BlockCol = #sparse_tensor.encoding<{63 map = (d0, d1) -> (d1 : compressed, d0 : dense)64}>65 66//67// Integration test that looks "under the hood" of sparse storage schemes.68//69module {70 //71 // Main driver that initializes a sparse tensor and inspects the sparse72 // storage schemes in detail. Note that users of the MLIR sparsifier73 // are typically not concerned with such details, but the test ensures74 // everything is working "under the hood".75 //76 func.func @main() {77 %c0 = arith.constant 0 : index78 %c1 = arith.constant 1 : index79 %d0 = arith.constant 0.0 : f6480 81 //82 // Initialize a dense tensor.83 //84 %t = arith.constant dense<[85 [ 1.0, 0.0, 2.0, 0.0, 0.0, 0.0, 0.0, 3.0],86 [ 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],87 [ 0.0, 0.0, 4.0, 0.0, 0.0, 0.0, 0.0, 0.0],88 [ 0.0, 0.0, 0.0, 5.0, 0.0, 0.0, 0.0, 0.0],89 [ 0.0, 0.0, 0.0, 0.0, 6.0, 0.0, 0.0, 0.0],90 [ 0.0, 7.0, 8.0, 0.0, 0.0, 0.0, 0.0, 9.0],91 [ 0.0, 0.0, 10.0, 0.0, 0.0, 0.0, 11.0, 12.0],92 [ 0.0, 13.0, 14.0, 0.0, 0.0, 0.0, 15.0, 16.0],93 [ 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],94 [ 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 17.0, 0.0]95 ]> : tensor<10x8xf64>96 97 //98 // Convert dense tensor to various sparse tensors.99 //100 %0 = sparse_tensor.convert %t : tensor<10x8xf64> to tensor<10x8xf64, #Dense>101 %1 = sparse_tensor.convert %t : tensor<10x8xf64> to tensor<10x8xf64, #CSR>102 %2 = sparse_tensor.convert %t : tensor<10x8xf64> to tensor<10x8xf64, #DCSR>103 %3 = sparse_tensor.convert %t : tensor<10x8xf64> to tensor<10x8xf64, #CSC>104 %4 = sparse_tensor.convert %t : tensor<10x8xf64> to tensor<10x8xf64, #DCSC>105 %x = sparse_tensor.convert %t : tensor<10x8xf64> to tensor<10x8xf64, #BlockRow>106 %y = sparse_tensor.convert %t : tensor<10x8xf64> to tensor<10x8xf64, #BlockCol>107 108 //109 // Inspect storage scheme of Dense.110 //111 // CHECK: ---- Sparse Tensor ----112 // CHECK-NEXT: nse = 80113 // CHECK-NEXT: dim = ( 10, 8 )114 // CHECK-NEXT: lvl = ( 10, 8 )115 // CHECK-NEXT: values : ( 1, 0, 2, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 0, 0, 0, 0, 6, 0, 0, 0, 0, 7, 8, 0, 0, 0, 0, 9, 0, 0, 10, 0, 0, 0, 11, 12, 0, 13, 14, 0, 0, 0, 15, 16, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 17, 0 )116 // CHECK-NEXT: ----117 //118 sparse_tensor.print %0 : tensor<10x8xf64, #Dense>119 120 //121 // Inspect storage scheme of CSR.122 //123 //124 // CHECK: ---- Sparse Tensor ----125 // CHECK-NEXT: nse = 17126 // CHECK-NEXT: dim = ( 10, 8 )127 // CHECK-NEXT: lvl = ( 10, 8 )128 // CHECK-NEXT: pos[1] : ( 0, 3, 3, 4, 5, 6, 9, 12, 16, 16, 17 )129 // CHECK-NEXT: crd[1] : ( 0, 2, 7, 2, 3, 4, 1, 2, 7, 2, 6, 7, 1, 2, 6, 7, 6 )130 // CHECK-NEXT: values : ( 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17 )131 // CHECK-NEXT: ----132 //133 sparse_tensor.print %1 : tensor<10x8xf64, #CSR>134 135 //136 // Inspect storage scheme of DCSR.137 //138 // CHECK: ---- Sparse Tensor ----139 // CHECK-NEXT: nse = 17140 // CHECK-NEXT: dim = ( 10, 8 )141 // CHECK-NEXT: lvl = ( 10, 8 )142 // CHECK-NEXT: pos[0] : ( 0, 8 )143 // CHECK-NEXT: crd[0] : ( 0, 2, 3, 4, 5, 6, 7, 9 )144 // CHECK-NEXT: pos[1] : ( 0, 3, 4, 5, 6, 9, 12, 16, 17 )145 // CHECK-NEXT: crd[1] : ( 0, 2, 7, 2, 3, 4, 1, 2, 7, 2, 6, 7, 1, 2, 6, 7, 6 )146 // CHECK-NEXT: values : ( 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17 )147 // CHECK-NEXT: ----148 //149 sparse_tensor.print %2 : tensor<10x8xf64, #DCSR>150 151 //152 // Inspect storage scheme of CSC.153 //154 // CHECK: ---- Sparse Tensor ----155 // CHECK-NEXT: nse = 17156 // CHECK-NEXT: dim = ( 10, 8 )157 // CHECK-NEXT: lvl = ( 8, 10 )158 // CHECK-NEXT: pos[1] : ( 0, 1, 3, 8, 9, 10, 10, 13, 17 )159 // CHECK-NEXT: crd[1] : ( 0, 5, 7, 0, 2, 5, 6, 7, 3, 4, 6, 7, 9, 0, 5, 6, 7 )160 // CHECK-NEXT: values : ( 1, 7, 13, 2, 4, 8, 10, 14, 5, 6, 11, 15, 17, 3, 9, 12, 16 )161 // CHECK-NEXT: ----162 //163 sparse_tensor.print %3 : tensor<10x8xf64, #CSC>164 165 //166 // Inspect storage scheme of DCSC.167 //168 // CHECK: ---- Sparse Tensor ----169 // CHECK-NEXT: nse = 17170 // CHECK-NEXT: dim = ( 10, 8 )171 // CHECK-NEXT: lvl = ( 8, 10 )172 // CHECK-NEXT: pos[0] : ( 0, 7 )173 // CHECK-NEXT: crd[0] : ( 0, 1, 2, 3, 4, 6, 7 )174 // CHECK-NEXT: pos[1] : ( 0, 1, 3, 8, 9, 10, 13, 17 )175 // CHECK-NEXT: crd[1] : ( 0, 5, 7, 0, 2, 5, 6, 7, 3, 4, 6, 7, 9, 0, 5, 6, 7 )176 // CHECK-NEXT: values : ( 1, 7, 13, 2, 4, 8, 10, 14, 5, 6, 11, 15, 17, 3, 9, 12, 16 )177 // CHECK-NEXT: ----178 //179 sparse_tensor.print %4 : tensor<10x8xf64, #DCSC>180 181 //182 // Inspect storage scheme of BlockRow.183 //184 // CHECK: ---- Sparse Tensor ----185 // CHECK-NEXT: nse = 64186 // CHECK-NEXT: dim = ( 10, 8 )187 // CHECK-NEXT: lvl = ( 10, 8 )188 // CHECK-NEXT: pos[0] : ( 0, 8 )189 // CHECK-NEXT: crd[0] : ( 0, 2, 3, 4, 5, 6, 7, 9 )190 // CHECK-NEXT: values : ( 1, 0, 2, 0, 0, 0, 0, 3, 0, 0, 4, 0, 0, 0, 0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 0, 0, 0, 0, 6, 0, 0, 0, 0, 7, 8, 0, 0, 0, 0, 9, 0, 0, 10, 0, 0, 0, 11, 12, 0, 13, 14, 0, 0, 0, 15, 16, 0, 0, 0, 0, 0, 0, 17, 0 )191 // CHECK-NEXT: ----192 //193 sparse_tensor.print %x : tensor<10x8xf64, #BlockRow>194 195 //196 // Inspect storage scheme of BlockCol.197 //198 // CHECK: ---- Sparse Tensor ----199 // CHECK-NEXT: nse = 70200 // CHECK-NEXT: dim = ( 10, 8 )201 // CHECK-NEXT: lvl = ( 8, 10 )202 // CHECK-NEXT: pos[0] : ( 0, 7 )203 // CHECK-NEXT: crd[0] : ( 0, 1, 2, 3, 4, 6, 7 )204 // CHECK-NEXT: values : ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7, 0, 13, 0, 0, 2, 0, 4, 0, 0, 8, 10, 14, 0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 15, 0, 17, 3, 0, 0, 0, 0, 9, 12, 16, 0, 0 )205 // CHECK-NEXT: ----206 //207 sparse_tensor.print %y : tensor<10x8xf64, #BlockCol>208 209 // Release the resources.210 bufferization.dealloc_tensor %0 : tensor<10x8xf64, #Dense>211 bufferization.dealloc_tensor %1 : tensor<10x8xf64, #CSR>212 bufferization.dealloc_tensor %2 : tensor<10x8xf64, #DCSR>213 bufferization.dealloc_tensor %3 : tensor<10x8xf64, #CSC>214 bufferization.dealloc_tensor %4 : tensor<10x8xf64, #DCSC>215 bufferization.dealloc_tensor %x : tensor<10x8xf64, #BlockRow>216 bufferization.dealloc_tensor %y : tensor<10x8xf64, #BlockCol>217 218 return219 }220}221