brintos

brintos / llvm-project-archived public Read only

0
0
Text · 5.0 KiB · 6e744e0 Raw
124 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// REDEFINE: %{env} = TENSOR0="%mlir_src_dir/test/Integration/data/test.tns"22// RUN: %{compile} | env %{env} %{run} | FileCheck %s23//24// Do the same run, but now with direct IR generation.25// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false26// RUN: %{compile} | env %{env} %{run} | FileCheck %s27//28// Do the same run, but now with direct IR generation and vectorization.29// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false vl=2 reassociate-fp-reductions=true enable-index-optimizations=true30// RUN: %{compile} | env %{env} %{run} | FileCheck %s31//32// Do the same run, but now with direct IR generation and VLA vectorization.33// RUN: %if mlir_arm_sve_tests %{ %{compile_sve} | env %{env} %{run_sve} | FileCheck %s %}34 35!Filename = !llvm.ptr36 37#SparseTensor = #sparse_tensor.encoding<{38  // Note that any dimToLvl permutation should give the same results39  // since, even though it impacts the sparse storage scheme layout,40  // it should not change the semantics.41  map = (d0, d1, d2, d3,42         d4, d5, d6, d7) -> (d7 : compressed, d6 : compressed,43                             d1 : compressed, d2 : compressed,44                             d0 : compressed, d3 : compressed,45                             d4 : compressed, d5 : compressed)46}>47 48#trait_flatten = {49  indexing_maps = [50    affine_map<(i,j,k,l,m,n,o,p) -> (i,j,k,l,m,n,o,p)>, // A51    affine_map<(i,j,k,l,m,n,o,p) -> (i,j)>              // X (out)52  ],53  iterator_types = [ "parallel",  "parallel",  "reduction", "reduction",54                     "reduction", "reduction", "reduction", "reduction" ],55  doc = "X(i,j) += A(i,j,k,l,m,n,o,p)"56}57 58//59// Integration test that lowers a kernel annotated as sparse to60// actual sparse code, initializes a matching sparse storage scheme61// from file, and runs the resulting code with the JIT compiler.62//63module {64  //65  // A kernel that flattens a rank 8 tensor into a dense matrix.66  //67  func.func @kernel_flatten(%arga: tensor<7x3x3x3x3x3x5x3xf64, #SparseTensor>,68                            %argx: tensor<7x3xf64>)69                                -> tensor<7x3xf64> {70    %0 = linalg.generic #trait_flatten71      ins(%arga: tensor<7x3x3x3x3x3x5x3xf64, #SparseTensor>)72      outs(%argx: tensor<7x3xf64>) {73      ^bb(%a: f64, %x: f64):74        %0 = arith.addf %x, %a : f6475        linalg.yield %0 : f6476    } -> tensor<7x3xf64>77    return %0 : tensor<7x3xf64>78  }79 80  func.func private @getTensorFilename(index) -> (!Filename)81  func.func private @printMemrefF64(%ptr : tensor<*xf64>)82 83  //84  // Main driver that reads tensor from file and calls the sparse kernel.85  //86  func.func @main() {87    %d0 = arith.constant 0.0 : f6488    %c0 = arith.constant 0 : index89    %c1 = arith.constant 1 : index90    %c3 = arith.constant 3 : index91    %c7 = arith.constant 7 : index92 93    // Setup matrix memory that is initialized to zero.94    %x = arith.constant dense<0.000000e+00> : tensor<7x3xf64>95 96    // Read the sparse tensor from file, construct sparse storage.97    %fileName = call @getTensorFilename(%c0) : (index) -> (!Filename)98    %a = sparse_tensor.new %fileName : !Filename to tensor<7x3x3x3x3x3x5x3xf64, #SparseTensor>99 100    // Call the kernel.101    %0 = call @kernel_flatten(%a, %x)102      : (tensor<7x3x3x3x3x3x5x3xf64, #SparseTensor>, tensor<7x3xf64>) -> tensor<7x3xf64>103 104    // Print the result for verification.105    //106    // CHECK:      {{\[}}[6.25,   0,   0],107    // CHECK-NEXT: [4.224,   6.21,   0],108    // CHECK-NEXT: [0,   0,   15.455],109    // CHECK-NEXT: [0,   0,   0],110    // CHECK-NEXT: [0,   0,   0],111    // CHECK-NEXT: [0,   0,   0],112    // CHECK-NEXT: [7,   0,   0]]113    //114    %1 = tensor.cast %0 : tensor<7x3xf64> to tensor<*xf64>115    call @printMemrefF64(%1) : (tensor<*xf64>) -> ()116 117    // Release the resources.118    bufferization.dealloc_tensor %a : tensor<7x3x3x3x3x3x5x3xf64, #SparseTensor>119    bufferization.dealloc_tensor %0 : tensor<7x3xf64>120 121    return122  }123}124