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#SparseVector = #sparse_tensor.encoding<{map = (d0) -> (d0 : compressed)}> 35 36#trait_op = { 37 indexing_maps = [ 38 affine_map<(i) -> (i)>, // a (in) 39 affine_map<(i) -> (i)>, // b (in) 40 affine_map<(i) -> (i)> // x (out) 41 ], 42 iterator_types = ["parallel"], 43 doc = "x(i) = a(i) OP b(i)" 44} 45 46module { 47 func.func @cadd(%arga: tensor<?xcomplex<f32>, #SparseVector>, 48 %argb: tensor<?xcomplex<f32>, #SparseVector>) 49 -> tensor<?xcomplex<f32>, #SparseVector> { 50 %c = arith.constant 0 : index 51 %d = tensor.dim %arga, %c : tensor<?xcomplex<f32>, #SparseVector> 52 %xv = tensor.empty(%d) : tensor<?xcomplex<f32>, #SparseVector> 53 %0 = linalg.generic #trait_op 54 ins(%arga, %argb: tensor<?xcomplex<f32>, #SparseVector>, 55 tensor<?xcomplex<f32>, #SparseVector>) 56 outs(%xv: tensor<?xcomplex<f32>, #SparseVector>) { 57 ^bb(%a: complex<f32>, %b: complex<f32>, %x: complex<f32>): 58 %1 = complex.add %a, %b : complex<f32> 59 linalg.yield %1 : complex<f32> 60 } -> tensor<?xcomplex<f32>, #SparseVector> 61 return %0 : tensor<?xcomplex<f32>, #SparseVector> 62 } 63 64 func.func @cmul(%arga: tensor<?xcomplex<f32>, #SparseVector>, 65 %argb: tensor<?xcomplex<f32>, #SparseVector>) 66 -> tensor<?xcomplex<f32>, #SparseVector> { 67 %c = arith.constant 0 : index 68 %d = tensor.dim %arga, %c : tensor<?xcomplex<f32>, #SparseVector> 69 %xv = tensor.empty(%d) : tensor<?xcomplex<f32>, #SparseVector> 70 %0 = linalg.generic #trait_op 71 ins(%arga, %argb: tensor<?xcomplex<f32>, #SparseVector>, 72 tensor<?xcomplex<f32>, #SparseVector>) 73 outs(%xv: tensor<?xcomplex<f32>, #SparseVector>) { 74 ^bb(%a: complex<f32>, %b: complex<f32>, %x: complex<f32>): 75 %1 = complex.mul %a, %b : complex<f32> 76 linalg.yield %1 : complex<f32> 77 } -> tensor<?xcomplex<f32>, #SparseVector> 78 return %0 : tensor<?xcomplex<f32>, #SparseVector> 79 } 80 81 // Driver method to call and verify complex kernels. 82 func.func @main() { 83 // Setup sparse vectors. 84 %v1 = arith.constant sparse< 85 [ [0], [28], [31] ], 86 [ (511.13, 2.0), (3.0, 4.0), (5.0, 6.0) ] > : tensor<32xcomplex<f32>> 87 %v2 = arith.constant sparse< 88 [ [1], [28], [31] ], 89 [ (1.0, 0.0), (2.0, 0.0), (3.0, 0.0) ] > : tensor<32xcomplex<f32>> 90 %sv1 = sparse_tensor.convert %v1 : tensor<32xcomplex<f32>> to tensor<?xcomplex<f32>, #SparseVector> 91 %sv2 = sparse_tensor.convert %v2 : tensor<32xcomplex<f32>> to tensor<?xcomplex<f32>, #SparseVector> 92 93 // Call sparse vector kernels. 94 %0 = call @cadd(%sv1, %sv2) 95 : (tensor<?xcomplex<f32>, #SparseVector>, 96 tensor<?xcomplex<f32>, #SparseVector>) -> tensor<?xcomplex<f32>, #SparseVector> 97 %1 = call @cmul(%sv1, %sv2) 98 : (tensor<?xcomplex<f32>, #SparseVector>, 99 tensor<?xcomplex<f32>, #SparseVector>) -> tensor<?xcomplex<f32>, #SparseVector> 100 101 // 102 // Verify the results. 103 // 104 // CHECK: ---- Sparse Tensor ---- 105 // CHECK-NEXT: nse = 4 106 // CHECK-NEXT: dim = ( 32 ) 107 // CHECK-NEXT: lvl = ( 32 ) 108 // CHECK-NEXT: pos[0] : ( 0, 4 ) 109 // CHECK-NEXT: crd[0] : ( 0, 1, 28, 31 ) 110 // CHECK-NEXT: values : ( ( 511.13, 2 ), ( 1, 0 ), ( 5, 4 ), ( 8, 6 ) ) 111 // CHECK-NEXT: ---- 112 // 113 // CHECK-NEXT: ---- Sparse Tensor ---- 114 // CHECK-NEXT: nse = 2 115 // CHECK-NEXT: dim = ( 32 ) 116 // CHECK-NEXT: lvl = ( 32 ) 117 // CHECK-NEXT: pos[0] : ( 0, 2 ) 118 // CHECK-NEXT: crd[0] : ( 28, 31 ) 119 // CHECK-NEXT: values : ( ( 6, 8 ), ( 15, 18 ) ) 120 // CHECK-NEXT: ---- 121 // 122 sparse_tensor.print %0 : tensor<?xcomplex<f32>, #SparseVector> 123 sparse_tensor.print %1 : tensor<?xcomplex<f32>, #SparseVector> 124 125 // Release the resources. 126 bufferization.dealloc_tensor %sv1 : tensor<?xcomplex<f32>, #SparseVector> 127 bufferization.dealloc_tensor %sv2 : tensor<?xcomplex<f32>, #SparseVector> 128 bufferization.dealloc_tensor %0 : tensor<?xcomplex<f32>, #SparseVector> 129 bufferization.dealloc_tensor %1 : tensor<?xcomplex<f32>, #SparseVector> 130 return 131 } 132} 133