103 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 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_mul_s = {37 indexing_maps = [38 affine_map<(i) -> (i)> // x (out)39 ],40 iterator_types = ["parallel"],41 doc = "x(i) = x(i) * 2.0"42}43 44module {45 func.func @main() {46 %f1 = arith.constant 1.0 : f3247 %f2 = arith.constant 2.0 : f3248 %f3 = arith.constant 3.0 : f3249 %f4 = arith.constant 4.0 : f3250 %c0 = arith.constant 0 : index51 %c1 = arith.constant 1 : index52 %c3 = arith.constant 3 : index53 %c8 = arith.constant 8 : index54 %c1023 = arith.constant 1023 : index55 56 // Build the sparse vector from straightline code.57 %0 = tensor.empty() : tensor<1024xf32, #SparseVector>58 %1 = tensor.insert %f1 into %0[%c0] : tensor<1024xf32, #SparseVector>59 %2 = tensor.insert %f2 into %1[%c1] : tensor<1024xf32, #SparseVector>60 %3 = tensor.insert %f3 into %2[%c3] : tensor<1024xf32, #SparseVector>61 %4 = tensor.insert %f4 into %3[%c1023] : tensor<1024xf32, #SparseVector>62 %5 = sparse_tensor.load %4 hasInserts : tensor<1024xf32, #SparseVector>63 64 //65 // CHECK: ---- Sparse Tensor ----66 // CHECK-NEXT: nse = 467 // CHECK-NEXT: dim = ( 1024 )68 // CHECK-NEXT: lvl = ( 1024 )69 // CHECK-NEXT: pos[0] : ( 0, 4 )70 // CHECK-NEXT: crd[0] : ( 0, 1, 3, 1023 )71 // CHECK-NEXT: values : ( 1, 2, 3, 4 )72 // CHECK-NEXT: ----73 //74 sparse_tensor.print %5 : tensor<1024xf32, #SparseVector>75 76 // Build another sparse vector in a loop.77 %6 = tensor.empty() : tensor<1024xf32, #SparseVector>78 %7 = scf.for %i = %c0 to %c8 step %c1 iter_args(%vin = %6) -> tensor<1024xf32, #SparseVector> {79 %ii = arith.muli %i, %c3 : index80 %vout = tensor.insert %f1 into %vin[%ii] : tensor<1024xf32, #SparseVector>81 scf.yield %vout : tensor<1024xf32, #SparseVector>82 }83 %8 = sparse_tensor.load %7 hasInserts : tensor<1024xf32, #SparseVector>84 85 //86 // CHECK-NEXT: ---- Sparse Tensor ----87 // CHECK-NEXT: nse = 888 // CHECK-NEXT: dim = ( 1024 )89 // CHECK-NEXT: lvl = ( 1024 )90 // CHECK-NEXT: pos[0] : ( 0, 8 )91 // CHECK-NEXT: crd[0] : ( 0, 3, 6, 9, 12, 15, 18, 21 )92 // CHECK-NEXT: values : ( 1, 1, 1, 1, 1, 1, 1, 1 )93 // CHECK-NEXT: ----94 //95 sparse_tensor.print %8 : tensor<1024xf32, #SparseVector>96 97 // Free resources.98 bufferization.dealloc_tensor %5 : tensor<1024xf32, #SparseVector>99 bufferization.dealloc_tensor %8 : tensor<1024xf32, #SparseVector>100 return101 }102}103