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 25// RUN: %{compile} | %{run} | FileCheck %s 26// 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=true 29// RUN: %{compile} | %{run} | FileCheck %s 30// 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#ST = #sparse_tensor.encoding<{map = (d0, d1, d2) -> (d0 : compressed, d1 : compressed, d2 : compressed)}> 35 36// 37// Trait for 3-d tensor element wise multiplication. 38// 39#trait_mul = { 40 indexing_maps = [ 41 affine_map<(i,j,k) -> (i,j,k)>, // A (in) 42 affine_map<(i,j,k) -> (i,j,k)>, // B (in) 43 affine_map<(i,j,k) -> (i,j,k)> // X (out) 44 ], 45 iterator_types = ["parallel", "parallel", "parallel"], 46 doc = "X(i,j,k) = A(i,j,k) * B(i,j,k)" 47} 48 49module { 50 // Multiplies two 3-d sparse tensors element-wise into a new sparse tensor. 51 func.func @tensor_mul(%arga: tensor<?x?x?xf64, #ST>, 52 %argb: tensor<?x?x?xf64, #ST>) -> tensor<?x?x?xf64, #ST> { 53 %c0 = arith.constant 0 : index 54 %c1 = arith.constant 1 : index 55 %c2 = arith.constant 2 : index 56 %d0 = tensor.dim %arga, %c0 : tensor<?x?x?xf64, #ST> 57 %d1 = tensor.dim %arga, %c1 : tensor<?x?x?xf64, #ST> 58 %d2 = tensor.dim %arga, %c2 : tensor<?x?x?xf64, #ST> 59 %xt = tensor.empty(%d0, %d1, %d2) : tensor<?x?x?xf64, #ST> 60 %0 = linalg.generic #trait_mul 61 ins(%arga, %argb: tensor<?x?x?xf64, #ST>, tensor<?x?x?xf64, #ST>) 62 outs(%xt: tensor<?x?x?xf64, #ST>) { 63 ^bb(%a: f64, %b: f64, %x: f64): 64 %1 = arith.mulf %a, %b : f64 65 linalg.yield %1 : f64 66 } -> tensor<?x?x?xf64, #ST> 67 return %0 : tensor<?x?x?xf64, #ST> 68 } 69 70 // Driver method to call and verify tensor multiplication kernel. 71 func.func @main() { 72 %c0 = arith.constant 0 : index 73 %default_val = arith.constant -1.0 : f64 74 75 // Setup sparse tensor A 76 %ta = arith.constant dense< 77 [ [ [1.0, 0.0, 0.0, 0.0, 0.0 ], 78 [0.0, 0.0, 0.0, 0.0, 0.0 ], 79 [1.2, 0.0, 3.5, 0.0, 0.0 ] ], 80 [ [0.0, 0.0, 0.0, 0.0, 0.0 ], 81 [0.0, 0.0, 0.0, 0.0, 0.0 ], 82 [0.0, 0.0, 0.0, 0.0, 0.0 ] ], 83 [ [2.0, 0.0, 0.0, 0.0, 0.0 ], 84 [0.0, 0.0, 0.0, 0.0, 0.0 ], 85 [0.0, 0.0, 4.0, 0.0, 0.0 ]] ]> : tensor<3x3x5xf64> 86 87 // Setup sparse tensor B 88 %tb = arith.constant dense< 89 [ [ [0.0, 0.0, 0.0, 0.0, 4.0 ], 90 [0.0, 0.0, 0.0, 0.0, 0.0 ], 91 [2.0, 0.0, 1.0, 0.0, 0.0 ] ], 92 [ [0.0, 0.0, 0.0, 0.0, 9.0 ], 93 [0.0, 0.0, 0.0, 0.0, 0.0 ], 94 [0.0, 7.0, 0.0, 0.0, 0.0 ] ], 95 [ [1.0, 0.0, 0.0, 0.0, 0.0 ], 96 [0.0, 0.0, 0.0, 0.0, 0.0 ], 97 [0.0, 0.0, 2.0, 0.0, 0.0 ]] ]> : tensor<3x3x5xf64> 98 99 %sta = sparse_tensor.convert %ta : tensor<3x3x5xf64> to tensor<?x?x?xf64, #ST> 100 %stb = sparse_tensor.convert %tb : tensor<3x3x5xf64> to tensor<?x?x?xf64, #ST> 101 102 103 // Call sparse tensor multiplication kernel. 104 %0 = call @tensor_mul(%sta, %stb) 105 : (tensor<?x?x?xf64, #ST>, tensor<?x?x?xf64, #ST>) -> tensor<?x?x?xf64, #ST> 106 107 // 108 // Verify results. 109 // 110 // CHECK: ---- Sparse Tensor ---- 111 // CHECK-NEXT: nse = 4 112 // CHECK-NEXT: dim = ( 3, 3, 5 ) 113 // CHECK-NEXT: lvl = ( 3, 3, 5 ) 114 // CHECK-NEXT: pos[0] : ( 0, 2 ) 115 // CHECK-NEXT: crd[0] : ( 0, 2 ) 116 // CHECK-NEXT: pos[1] : ( 0, 1, 3 ) 117 // CHECK-NEXT: crd[1] : ( 2, 0, 2 ) 118 // CHECK-NEXT: pos[2] : ( 0, 2, 3, 4 ) 119 // CHECK-NEXT: crd[2] : ( 0, 2, 0, 2 ) 120 // CHECK-NEXT: values : ( 2.4, 3.5, 2, 8 ) 121 // CHECK-NEXT: ---- 122 // 123 sparse_tensor.print %0 : tensor<?x?x?xf64, #ST> 124 125 // Release the resources. 126 bufferization.dealloc_tensor %sta : tensor<?x?x?xf64, #ST> 127 bufferization.dealloc_tensor %stb : tensor<?x?x?xf64, #ST> 128 bufferization.dealloc_tensor %0 : tensor<?x?x?xf64, #ST> 129 130 return 131 } 132} 133