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
|
//===- StructuredOpsUtils.cpp - Utilities used by structured ops ----------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include "mlir/Dialect/Utils/StructuredOpsUtils.h"
#include "mlir/IR/AffineMap.h"
#include "mlir/IR/Builders.h"
#include "mlir/IR/BuiltinAttributes.h"
#include "mlir/IR/IRMapping.h"
#include "mlir/Dialect/Utils/DialectUtilsEnums.cpp.inc"
using namespace mlir;
bool mlir::isRowMajorMatmul(ArrayAttr indexingMaps) {
if (indexingMaps.size() != 3)
return false;
auto map0 = indexingMaps[0].cast<AffineMapAttr>().getValue();
auto map1 = indexingMaps[1].cast<AffineMapAttr>().getValue();
auto map2 = indexingMaps[2].cast<AffineMapAttr>().getValue();
if (map0.getNumResults() != 2 || map1.getNumResults() != 2 ||
map2.getNumResults() != 2 || map0.getNumInputs() != 3 ||
map1.getNumInputs() != 3 || map2.getNumInputs() != 3) {
return false;
}
// Extract dimensions for MxK * KxN -> MxN
AffineExpr m = map2.getResult(0);
AffineExpr n = map2.getResult(1);
AffineExpr k = map0.getResult(1);
auto *context = indexingMaps.getContext();
auto mapA = AffineMapAttr::get(AffineMap::get(3, 0, {m, k}, context));
auto mapB = AffineMapAttr::get(AffineMap::get(3, 0, {k, n}, context));
auto mapC = AffineMapAttr::get(AffineMap::get(3, 0, {m, n}, context));
auto maps = ArrayAttr::get(context, {mapA, mapB, mapC});
return indexingMaps == maps;
}
bool mlir::isColumnMajorMatmul(ArrayAttr indexingMaps) {
if (indexingMaps.size() != 3)
return false;
auto map0 = indexingMaps[0].cast<AffineMapAttr>().getValue();
auto map1 = indexingMaps[1].cast<AffineMapAttr>().getValue();
auto map2 = indexingMaps[2].cast<AffineMapAttr>().getValue();
if (map0.getNumResults() != 2 || map1.getNumResults() != 2 ||
map2.getNumResults() != 2 || map0.getNumInputs() != 3 ||
map1.getNumInputs() != 3 || map2.getNumInputs() != 3) {
return false;
}
// Extract dimensions for KxM * NxK -> NxM
AffineExpr n = map2.getResult(0);
AffineExpr m = map2.getResult(1);
AffineExpr k = map0.getResult(0);
auto *context = indexingMaps.getContext();
auto mapA = AffineMapAttr::get(AffineMap::get(3, 0, {k, m}, context));
auto mapB = AffineMapAttr::get(AffineMap::get(3, 0, {n, k}, context));
auto mapC = AffineMapAttr::get(AffineMap::get(3, 0, {n, m}, context));
auto maps = ArrayAttr::get(context, {mapA, mapB, mapC});
return indexingMaps == maps;
}
bool mlir::isRowMajorBatchMatmul(ArrayAttr indexingMaps) {
if (indexingMaps.size() != 3)
return false;
auto map0 = indexingMaps[0].cast<AffineMapAttr>().getValue();
auto map1 = indexingMaps[1].cast<AffineMapAttr>().getValue();
auto map2 = indexingMaps[2].cast<AffineMapAttr>().getValue();
if (map0.getNumResults() != 3 || map1.getNumResults() != 3 ||
map2.getNumResults() != 3 || map0.getNumInputs() != 4 ||
map1.getNumInputs() != 4 || map2.getNumInputs() != 4) {
return false;
}
// Extract dimensions for BxMxK * BxKxN -> BxMxN
AffineExpr b = map2.getResult(0);
AffineExpr m = map2.getResult(1);
AffineExpr n = map2.getResult(2);
AffineExpr k = map0.getResult(2);
auto *context = indexingMaps.getContext();
auto mapA = AffineMapAttr::get(AffineMap::get(4, 0, {b, m, k}, context));
auto mapB = AffineMapAttr::get(AffineMap::get(4, 0, {b, k, n}, context));
auto mapC = AffineMapAttr::get(AffineMap::get(4, 0, {b, m, n}, context));
auto maps = ArrayAttr::get(context, {mapA, mapB, mapC});
return indexingMaps == maps;
}
Operation *mlir::clone(OpBuilder &b, Operation *op, TypeRange newResultTypes,
ValueRange newOperands) {
IRMapping bvm;
OperationState state(op->getLoc(), op->getName(), newOperands, newResultTypes,
op->getAttrs());
for (Region &r : op->getRegions())
r.cloneInto(state.addRegion(), bvm);
return b.create(state);
}
Operation *mlir::cloneWithoutRegions(OpBuilder &b, Operation *op,
TypeRange newResultTypes,
ValueRange newOperands) {
OperationState state(op->getLoc(), op->getName(), newOperands, newResultTypes,
op->getAttrs());
for (size_t cnt = 0, e = op->getNumRegions(); cnt < e; ++cnt)
state.addRegion();
return b.create(state);
}
|
