xref: /llvm-project/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_sum_f16.mlir (revision eb206e9ea84eff0a0596fed2de8316d924f946d1)
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!Filename = !llvm.ptr
35
36#SparseMatrix = #sparse_tensor.encoding<{
37  map = (d0, d1) -> (d0 : compressed, d1 : compressed)
38}>
39
40#trait_sum_reduce = {
41  indexing_maps = [
42    affine_map<(i,j) -> (i,j)>, // A
43    affine_map<(i,j) -> ()>     // x (out)
44  ],
45  iterator_types = ["reduction", "reduction"],
46  doc = "x += A(i,j)"
47}
48
49module {
50  //
51  // A kernel that sum-reduces a matrix to a single scalar.
52  //
53  func.func @kernel_sum_reduce(%arga: tensor<?x?xf16, #SparseMatrix>,
54                               %argx: tensor<f16>) -> tensor<f16> {
55    %0 = linalg.generic #trait_sum_reduce
56      ins(%arga: tensor<?x?xf16, #SparseMatrix>)
57      outs(%argx: tensor<f16>) {
58      ^bb(%a: f16, %x: f16):
59        %0 = arith.addf %x, %a : f16
60        linalg.yield %0 : f16
61    } -> tensor<f16>
62    return %0 : tensor<f16>
63  }
64
65  func.func private @getTensorFilename(index) -> (!Filename)
66
67  //
68  // Main driver that reads matrix from file and calls the sparse kernel.
69  //
70  func.func @main() {
71    // Setup input sparse matrix from compressed constant.
72    %d = arith.constant dense <[
73       [ 1.1,  1.2,  0.0,  1.4 ],
74       [ 0.0,  0.0,  0.0,  0.0 ],
75       [ 3.1,  0.0,  3.3,  3.4 ]
76    ]> : tensor<3x4xf16>
77    %a = sparse_tensor.convert %d : tensor<3x4xf16> to tensor<?x?xf16, #SparseMatrix>
78
79    %d0 = arith.constant 0.0 : f16
80    // Setup memory for a single reduction scalar,
81    // initialized to zero.
82    %x = tensor.from_elements %d0 : tensor<f16>
83
84    // Call the kernel.
85    %0 = call @kernel_sum_reduce(%a, %x)
86      : (tensor<?x?xf16, #SparseMatrix>, tensor<f16>) -> tensor<f16>
87
88    // Print the result for verification.
89    //
90    // CHECK: 13.5
91    //
92    %v = tensor.extract %0[] : tensor<f16>
93    %vf = arith.extf %v: f16 to f32
94    vector.print %vf : f32
95
96    // Release the resources.
97    bufferization.dealloc_tensor %0 : tensor<f16>
98    bufferization.dealloc_tensor %a : tensor<?x?xf16, #SparseMatrix>
99
100    return
101  }
102}
103