brintos

brintos / llvm-project-archived public Read only

0
0
Text · 5.9 KiB · 77450a3 Raw
149 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=false enable-buffer-initialization=true25// RUN: %{compile} | %{run} | FileCheck %s26//27// Do the same run, but now with direct IR generation and 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 direct IR generation and VLA vectorization.32// RUN: %if mlir_arm_sve_tests %{ %{compile_sve} | %{run_sve} | FileCheck %s %}33 34#SparseVector = #sparse_tensor.encoding<{ map = (d0) -> (d0 : compressed) }>35 36#trait_op = {37  indexing_maps = [38    affine_map<(i) -> (i)>, // a39    affine_map<(i) -> (i)>  // x (out)40  ],41  iterator_types = ["parallel"],42  doc = "x(i) = OP a(i)"43}44 45module {46  func.func @sparse_absf(%arg0: tensor<?xf64, #SparseVector>)47                             -> tensor<?xf64, #SparseVector> {48    %c0 = arith.constant 0 : index49    %d = tensor.dim %arg0, %c0 : tensor<?xf64, #SparseVector>50    %xin = tensor.empty(%d) : tensor<?xf64, #SparseVector>51    %0 = linalg.generic #trait_op52      ins(%arg0: tensor<?xf64, #SparseVector>)53      outs(%xin: tensor<?xf64, #SparseVector>) {54      ^bb0(%a: f64, %x: f64) :55        %result = math.absf %a : f6456        linalg.yield %result : f6457    } -> tensor<?xf64, #SparseVector>58    return %0 : tensor<?xf64, #SparseVector>59  }60 61  func.func @sparse_absi(%arg0: tensor<?xi32, #SparseVector>)62                             -> tensor<?xi32, #SparseVector> {63    %c0 = arith.constant 0 : index64    %d = tensor.dim %arg0, %c0 : tensor<?xi32, #SparseVector>65    %xin = tensor.empty(%d) : tensor<?xi32, #SparseVector>66    %0 = linalg.generic #trait_op67      ins(%arg0: tensor<?xi32, #SparseVector>)68      outs(%xin: tensor<?xi32, #SparseVector>) {69      ^bb0(%a: i32, %x: i32) :70        %result = math.absi %a : i3271        linalg.yield %result : i3272    } -> tensor<?xi32, #SparseVector>73    return %0 : tensor<?xi32, #SparseVector>74  }75 76  // Driver method to call and verify sign kernel.77  func.func @main() {78    %c0 = arith.constant 0 : index79    %df = arith.constant 99.99 : f6480    %di = arith.constant 9999 : i3281 82    %pnan = arith.constant 0x7FF0000001000000 : f6483    %nnan = arith.constant 0xFFF0000001000000 : f6484    %pinf = arith.constant 0x7FF0000000000000 : f6485    %ninf = arith.constant 0xFFF0000000000000 : f6486 87    // Setup sparse vectors.88    %v1 = arith.constant sparse<89       [ [0], [3], [5], [11], [13], [17], [18], [20], [21], [28], [29], [31] ],90         [ -1.5, 1.5, -10.2, 11.3, 1.0, -1.0,91           0x7FF0000001000000, // +NaN92           0xFFF0000001000000, // -NaN93           0x7FF0000000000000, // +Inf94           0xFFF0000000000000, // -Inf95           -0.0,               // -Zero96           0.0                 // +Zero97        ]98    > : tensor<32xf64>99    %v2 = arith.constant sparse<100       [ [0], [3], [5], [11], [13], [17], [18], [21], [31] ],101         [ -2147483648, -2147483647, -1000, -1, 0,102           1, 1000, 2147483646, 2147483647103         ]104    > : tensor<32xi32>105    %sv1 = sparse_tensor.convert %v1106         : tensor<32xf64> to tensor<?xf64, #SparseVector>107    %sv2 = sparse_tensor.convert %v2108         : tensor<32xi32> to tensor<?xi32, #SparseVector>109 110    // Call abs kernels.111    %0 = call @sparse_absf(%sv1) : (tensor<?xf64, #SparseVector>)112                                 -> tensor<?xf64, #SparseVector>113 114    %1 = call @sparse_absi(%sv2) : (tensor<?xi32, #SparseVector>)115                                 -> tensor<?xi32, #SparseVector>116 117    //118    // Verify the results.119    //120    // CHECK:      ---- Sparse Tensor ----121    // CHECK-NEXT: nse = 12122    // CHECK-NEXT: dim = ( 32 )123    // CHECK-NEXT: lvl = ( 32 )124    // CHECK-NEXT: pos[0] : ( 0, 12 )125    // CHECK-NEXT: crd[0] : ( 0, 3, 5, 11, 13, 17, 18, 20, 21, 28, 29, 31 )126    // CHECK-NEXT: values : ( 1.5, 1.5, 10.2, 11.3, 1, 1, nan, nan, inf, inf, 0, 0 )127    // CHECK-NEXT: ----128    //129    // CHECK-NEXT: ---- Sparse Tensor ----130    // CHECK-NEXT: nse = 9131    // CHECK-NEXT: dim = ( 32 )132    // CHECK-NEXT: lvl = ( 32 )133    // CHECK-NEXT: pos[0] : ( 0, 9 )134    // CHECK-NEXT: crd[0] : ( 0, 3, 5, 11, 13, 17, 18, 21, 31 )135    // CHECK-NEXT: values : ( -2147483648, 2147483647, 1000, 1, 0, 1, 1000, 2147483646, 2147483647 )136    // CHECK-NEXT: ----137    //138    sparse_tensor.print %0 : tensor<?xf64, #SparseVector>139    sparse_tensor.print %1 : tensor<?xi32, #SparseVector>140 141    // Release the resources.142    bufferization.dealloc_tensor %sv1 : tensor<?xf64, #SparseVector>143    bufferization.dealloc_tensor %sv2 : tensor<?xi32, #SparseVector>144    bufferization.dealloc_tensor %0 : tensor<?xf64, #SparseVector>145    bufferization.dealloc_tensor %1 : tensor<?xi32, #SparseVector>146    return147  }148}149