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 35#map = affine_map<(d0) -> (d0)> 36 37#SV = #sparse_tensor.encoding<{ 38 map = (d0) -> (d0 : compressed) 39}> 40 41module { 42 43 // This directly yields an empty sparse vector. 44 func.func @empty() -> tensor<10xf32, #SV> { 45 %0 = tensor.empty() : tensor<10xf32, #SV> 46 return %0 : tensor<10xf32, #SV> 47 } 48 49 // This also directly yields an empty sparse vector. 50 func.func @empty_alloc() -> tensor<10xf32, #SV> { 51 %0 = bufferization.alloc_tensor() : tensor<10xf32, #SV> 52 return %0 : tensor<10xf32, #SV> 53 } 54 55 // This yields a hidden empty sparse vector (all zeros). 56 func.func @zeros() -> tensor<10xf32, #SV> { 57 %cst = arith.constant 0.0 : f32 58 %0 = bufferization.alloc_tensor() : tensor<10xf32, #SV> 59 %1 = linalg.generic { 60 indexing_maps = [#map], 61 iterator_types = ["parallel"]} 62 outs(%0 : tensor<10xf32, #SV>) { 63 ^bb0(%out: f32): 64 linalg.yield %cst : f32 65 } -> tensor<10xf32, #SV> 66 return %1 : tensor<10xf32, #SV> 67 } 68 69 // This yields a filled sparse vector (all ones). 70 func.func @ones() -> tensor<10xf32, #SV> { 71 %cst = arith.constant 1.0 : f32 72 %0 = bufferization.alloc_tensor() : tensor<10xf32, #SV> 73 %1 = linalg.generic { 74 indexing_maps = [#map], 75 iterator_types = ["parallel"]} 76 outs(%0 : tensor<10xf32, #SV>) { 77 ^bb0(%out: f32): 78 linalg.yield %cst : f32 79 } -> tensor<10xf32, #SV> 80 return %1 : tensor<10xf32, #SV> 81 } 82 83 // 84 // Main driver. 85 // 86 func.func @main() { 87 88 %0 = call @empty() : () -> tensor<10xf32, #SV> 89 %1 = call @empty_alloc() : () -> tensor<10xf32, #SV> 90 %2 = call @zeros() : () -> tensor<10xf32, #SV> 91 %3 = call @ones() : () -> tensor<10xf32, #SV> 92 93 // 94 // Verify the output. In particular, make sure that 95 // all empty sparse vector data structures are properly 96 // finalized with a pair (0,0) for positions. 97 // 98 // CHECK: ---- Sparse Tensor ---- 99 // CHECK-NEXT: nse = 0 100 // CHECK-NEXT: dim = ( 10 ) 101 // CHECK-NEXT: lvl = ( 10 ) 102 // CHECK-NEXT: pos[0] : ( 0, 0 ) 103 // CHECK-NEXT: crd[0] : ( ) 104 // CHECK-NEXT: values : ( ) 105 // CHECK-NEXT: ---- 106 // 107 // CHECK-NEXT: ---- Sparse Tensor ---- 108 // CHECK-NEXT: nse = 0 109 // CHECK-NEXT: dim = ( 10 ) 110 // CHECK-NEXT: lvl = ( 10 ) 111 // CHECK-NEXT: pos[0] : ( 0, 0 ) 112 // CHECK-NEXT: crd[0] : ( ) 113 // CHECK-NEXT: values : ( ) 114 // CHECK-NEXT: ---- 115 // 116 // CHECK-NEXT: ---- Sparse Tensor ---- 117 // CHECK-NEXT: nse = 0 118 // CHECK-NEXT: dim = ( 10 ) 119 // CHECK-NEXT: lvl = ( 10 ) 120 // CHECK-NEXT: pos[0] : ( 0, 0 ) 121 // CHECK-NEXT: crd[0] : ( ) 122 // CHECK-NEXT: values : ( ) 123 // CHECK-NEXT: ---- 124 // 125 // CHECK-NEXT: ---- Sparse Tensor ---- 126 // CHECK-NEXT: nse = 10 127 // CHECK-NEXT: dim = ( 10 ) 128 // CHECK-NEXT: lvl = ( 10 ) 129 // CHECK-NEXT: pos[0] : ( 0, 10 ) 130 // CHECK-NEXT: crd[0] : ( 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 ) 131 // CHECK-NEXT: values : ( 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 ) 132 // CHECK-NEXT: ---- 133 // 134 sparse_tensor.print %0 : tensor<10xf32, #SV> 135 sparse_tensor.print %1 : tensor<10xf32, #SV> 136 sparse_tensor.print %2 : tensor<10xf32, #SV> 137 sparse_tensor.print %3 : tensor<10xf32, #SV> 138 139 bufferization.dealloc_tensor %0 : tensor<10xf32, #SV> 140 bufferization.dealloc_tensor %1 : tensor<10xf32, #SV> 141 bufferization.dealloc_tensor %2 : tensor<10xf32, #SV> 142 bufferization.dealloc_tensor %3 : tensor<10xf32, #SV> 143 return 144 } 145} 146