1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
|
//--------------------------------------------------------------------------------------------------
// WHEN CREATING A NEW TEST, PLEASE JUST COPY & PASTE WITHOUT EDITS.
//
// Set-up that's shared across all tests in this directory. In principle, this
// config could be moved to lit.local.cfg. However, there are downstream users that
// do not use these LIT config files. Hence why this is kept inline.
//
// DEFINE: %{sparsifier_opts} = enable-runtime-library=true
// DEFINE: %{sparsifier_opts_sve} = enable-arm-sve=true %{sparsifier_opts}
// DEFINE: %{compile} = mlir-opt %s --sparsifier="%{sparsifier_opts}"
// DEFINE: %{compile_sve} = mlir-opt %s --sparsifier="%{sparsifier_opts_sve}"
// DEFINE: %{run_libs} = -shared-libs=%mlir_c_runner_utils,%mlir_runner_utils
// DEFINE: %{run_libs_sve} = -shared-libs=%native_mlir_runner_utils,%native_mlir_c_runner_utils
// DEFINE: %{run_opts} = -e main -entry-point-result=void
// DEFINE: %{run} = mlir-runner %{run_opts} %{run_libs}
// DEFINE: %{run_sve} = %mcr_aarch64_cmd --march=aarch64 --mattr="+sve" %{run_opts} %{run_libs_sve}
//
// DEFINE: %{env} =
//--------------------------------------------------------------------------------------------------
// REDEFINE: %{env} = TENSOR0="%mlir_src_dir/test/Integration/data/wide.mtx"
// RUN: %{compile} | env %{env} %{run} | FileCheck %s
//
// Do the same run, but now with direct IR generation.
// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false
// RUN: %{compile} | env %{env} %{run} | FileCheck %s
//
// Do the same run, but now with parallelization strategy.
// REDEFINE: %{sparsifier_opts} = enable-runtime-library=true parallelization-strategy=any-storage-any-loop
// RUN: %{compile} | env %{env} %{run} | FileCheck %s
//
// Do the same run, but now with direct IR generation and parallelization strategy.
// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false parallelization-strategy=any-storage-any-loop
// RUN: %{compile} | env %{env} %{run} | FileCheck %s
//
// Do the same run, but now with direct IR generation and vectorization.
// REDEFINE: %{sparsifier_opts} = enable-runtime-library=false vl=2 reassociate-fp-reductions=true enable-index-optimizations=true
// RUN: %{compile} | env %{env} %{run} | FileCheck %s
//
// Do the same run, but now with direct IR generation and, if available, VLA
// vectorization.
// RUN: %if mlir_arm_sve_tests %{ %{compile_sve} | env %{env} %{run_sve} | FileCheck %s %}
!Filename = !llvm.ptr
#SparseMatrix = #sparse_tensor.encoding<{
map = (d0, d1) -> (d0 : dense, d1 : compressed),
posWidth = 8,
crdWidth = 8
}>
#matvec = {
indexing_maps = [
affine_map<(i,j) -> (i,j)>, // A
affine_map<(i,j) -> (j)>, // b
affine_map<(i,j) -> (i)> // x (out)
],
iterator_types = ["parallel", "reduction"],
doc = "X(i) += A(i,j) * B(j)"
}
//
// Integration test that lowers a kernel annotated as sparse to
// actual sparse code, initializes a matching sparse storage scheme
// from file, and runs the resulting code with the JIT compiler.
//
module {
//
// A kernel that multiplies a sparse matrix A with a dense vector b
// into a dense vector x.
//
func.func @kernel_matvec(%arga: tensor<?x?xi32, #SparseMatrix>,
%argb: tensor<?xi32>,
%argx: tensor<?xi32>)
-> tensor<?xi32> {
%0 = linalg.generic #matvec
ins(%arga, %argb: tensor<?x?xi32, #SparseMatrix>, tensor<?xi32>)
outs(%argx: tensor<?xi32>) {
^bb(%a: i32, %b: i32, %x: i32):
%0 = arith.muli %a, %b : i32
%1 = arith.addi %x, %0 : i32
linalg.yield %1 : i32
} -> tensor<?xi32>
return %0 : tensor<?xi32>
}
func.func private @getTensorFilename(index) -> (!Filename)
//
// Main driver that reads matrix from file and calls the sparse kernel.
//
func.func @main() {
%i0 = arith.constant 0 : i32
%c0 = arith.constant 0 : index
%c1 = arith.constant 1 : index
%c4 = arith.constant 4 : index
%c256 = arith.constant 256 : index
// Read the sparse matrix from file, construct sparse storage.
%fileName = call @getTensorFilename(%c0) : (index) -> (!Filename)
%a = sparse_tensor.new %fileName : !Filename to tensor<?x?xi32, #SparseMatrix>
// Initialize dense vectors.
%b = tensor.generate %c256 {
^bb0(%i : index):
%k = arith.addi %i, %c1 : index
%j = arith.index_cast %k : index to i32
tensor.yield %j : i32
} : tensor<?xi32>
%x = tensor.generate %c4 {
^bb0(%i : index):
tensor.yield %i0 : i32
} : tensor<?xi32>
// Call kernel.
%0 = call @kernel_matvec(%a, %b, %x)
: (tensor<?x?xi32, #SparseMatrix>, tensor<?xi32>, tensor<?xi32>) -> tensor<?xi32>
// Print the result for verification.
//
// CHECK: ( 889, 1514, -21, -3431 )
//
%v = vector.transfer_read %0[%c0], %i0: tensor<?xi32>, vector<4xi32>
vector.print %v : vector<4xi32>
// Release the resources.
bufferization.dealloc_tensor %a : tensor<?x?xi32, #SparseMatrix>
bufferization.dealloc_tensor %b : tensor<?xi32>
bufferization.dealloc_tensor %0 : tensor<?xi32>
return
}
}
|
