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
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
|
# RUN: %PYTHON %s 2>&1 | FileCheck %s
from mlir.passmanager import PassManager
from mlir.ir import Context, Location, Module, InsertionPoint, UnitAttr
from mlir.dialects import scf, pdl, func, arith, linalg
from mlir.dialects.transform import (
get_parent_op,
apply_patterns_canonicalization,
apply_cse,
any_op_t,
)
from mlir.dialects.transform.structured import structured_match
from mlir.dialects.transform.loop import loop_unroll
from mlir.dialects.transform.extras import named_sequence, apply_patterns
from mlir.extras import types as T
from mlir.dialects.builtin import module, ModuleOp
def construct_and_print_in_module(f):
print("\nTEST:", f.__name__)
with Context(), Location.unknown():
module = Module.create()
with InsertionPoint(module.body):
module = f(module)
if module is not None:
print(module)
return f
# CHECK-LABEL: TEST: test_named_sequence
@construct_and_print_in_module
def test_named_sequence(module_):
# CHECK-LABEL: func.func @loop_unroll_op() {
# CHECK: %[[VAL_0:.*]] = arith.constant 0 : index
# CHECK: %[[VAL_1:.*]] = arith.constant 42 : index
# CHECK: %[[VAL_2:.*]] = arith.constant 5 : index
# CHECK: scf.for %[[VAL_3:.*]] = %[[VAL_0]] to %[[VAL_1]] step %[[VAL_2]] {
# CHECK: %[[VAL_4:.*]] = arith.addi %[[VAL_3]], %[[VAL_3]] : index
# CHECK: }
# CHECK: return
# CHECK: }
@func.func()
def loop_unroll_op():
for i in scf.for_(0, 42, 5):
v = arith.addi(i, i)
scf.yield_([])
# CHECK-LABEL: module attributes {transform.with_named_sequence} {
# CHECK: transform.named_sequence @__transform_main(%[[VAL_0:.*]]: !transform.any_op) {
# CHECK: %[[VAL_1:.*]] = transform.structured.match ops{["arith.addi"]} in %[[VAL_0]] : (!transform.any_op) -> !transform.any_op
# CHECK: %[[VAL_2:.*]] = transform.get_parent_op %[[VAL_1]] {op_name = "scf.for"} : (!transform.any_op) -> !pdl.operation
# CHECK: transform.loop.unroll %[[VAL_2]] {factor = 4 : i64} : !pdl.operation
# CHECK: transform.yield
# CHECK: }
# CHECK: }
@module(attrs={"transform.with_named_sequence": UnitAttr.get()})
def mod():
@named_sequence("__transform_main", [any_op_t()], [])
def basic(target: any_op_t()):
m = structured_match(any_op_t(), target, ops=["arith.addi"])
loop = get_parent_op(pdl.op_t(), m, op_name="scf.for")
loop_unroll(loop, 4)
# The identifier (name) of the function becomes the Operation
assert isinstance(mod.opview, ModuleOp)
print(module_)
pm = PassManager.parse("builtin.module(transform-interpreter)")
pm.run(module_.operation)
# CHECK-LABEL: func.func @loop_unroll_op() {
# CHECK: %[[VAL_0:.*]] = arith.constant 0 : index
# CHECK: %[[VAL_1:.*]] = arith.constant 42 : index
# CHECK: %[[VAL_2:.*]] = arith.constant 5 : index
# CHECK: %[[VAL_6:.*]] = arith.constant 40 : index
# CHECK: %[[VAL_7:.*]] = arith.constant 20 : index
# CHECK: scf.for %[[VAL_3:.*]] = %[[VAL_0]] to %[[VAL_6]] step %[[VAL_7]] {
# CHECK: %[[VAL_5:.*]] = arith.addi %[[VAL_3]], %[[VAL_3]] : index
# CHECK: %[[VAL_8:.*]] = arith.constant 1 : index
# CHECK: %[[VAL_9:.*]] = arith.muli %[[VAL_2]], %[[VAL_8]] : index
# CHECK: %[[VAL_10:.*]] = arith.addi %[[VAL_3]], %[[VAL_9]] : index
# CHECK: %[[VAL_11:.*]] = arith.addi %[[VAL_10]], %[[VAL_10]] : index
# CHECK: %[[VAL_12:.*]] = arith.constant 2 : index
# CHECK: %[[VAL_13:.*]] = arith.muli %[[VAL_2]], %[[VAL_12]] : index
# CHECK: %[[VAL_14:.*]] = arith.addi %[[VAL_3]], %[[VAL_13]] : index
# CHECK: %[[VAL_15:.*]] = arith.addi %[[VAL_14]], %[[VAL_14]] : index
# CHECK: %[[VAL_16:.*]] = arith.constant 3 : index
# CHECK: %[[VAL_17:.*]] = arith.muli %[[VAL_2]], %[[VAL_16]] : index
# CHECK: %[[VAL_18:.*]] = arith.addi %[[VAL_3]], %[[VAL_17]] : index
# CHECK: %[[VAL_19:.*]] = arith.addi %[[VAL_18]], %[[VAL_18]] : index
# CHECK: }
# CHECK: %[[VAL_4:.*]] = arith.addi %[[VAL_6]], %[[VAL_6]] : index
# CHECK: return
# CHECK: }
print(module_)
# CHECK-LABEL: TEST: test_apply_patterns
@construct_and_print_in_module
def test_apply_patterns(module_):
b, M, N, K = 1, 3, 5, 3
# CHECK-LABEL: func.func @batch_reduce_matmul(
# CHECK-SAME: %[[VAL_0:.*]]: tensor<1x3x5xf32>,
# CHECK-SAME: %[[VAL_1:.*]]: tensor<1x5x3xf32>,
# CHECK-SAME: %[[VAL_2:.*]]: tensor<3x3xf32>) -> tensor<3x3xf32> {
# CHECK: %[[VAL_3:.*]] = arith.constant 1 : i32
# CHECK: %[[VAL_4:.*]] = arith.addi %[[VAL_3]], %[[VAL_3]] : i32
# CHECK: %[[VAL_5:.*]] = linalg.batch_reduce_matmul ins(%[[VAL_0]], %[[VAL_1]] : tensor<1x3x5xf32>, tensor<1x5x3xf32>) outs(%[[VAL_2]] : tensor<3x3xf32>) -> tensor<3x3xf32>
# CHECK: return %[[VAL_5]] : tensor<3x3xf32>
# CHECK: }
@func.func(
T.tensor(b, M, N, T.f32()), T.tensor(b, N, K, T.f32()), T.tensor(M, K, T.f32())
)
def batch_reduce_matmul(A, B, C):
i = arith.constant(T.i32(), 1)
v = arith.addi(i, i)
return linalg.batch_reduce_matmul(A, B, outs=[C])
# CHECK-LABEL: module attributes {transform.with_named_sequence} {
# CHECK: transform.named_sequence @__transform_main(%[[VAL_0:.*]]: !transform.any_op) {
# CHECK: %[[VAL_1:.*]] = transform.structured.match ops{["linalg.batch_reduce_matmul"]} in %[[VAL_0]] : (!transform.any_op) -> !transform.any_op
# CHECK: %[[VAL_2:.*]] = transform.get_parent_op %[[VAL_1]] {op_name = "func.func"} : (!transform.any_op) -> !pdl.operation
# CHECK: transform.apply_patterns to %[[VAL_2]] {
# CHECK: transform.apply_patterns.canonicalization
# CHECK: } : !pdl.operation
# CHECK: %[[VAL_3:.*]] = transform.structured.match ops{["func.func"]} in %[[VAL_0]] : (!transform.any_op) -> !transform.any_op
# CHECK: transform.apply_cse to %[[VAL_3]] : !transform.any_op
# CHECK: transform.yield
# CHECK: }
# CHECK: }
@module(attrs={"transform.with_named_sequence": UnitAttr.get()})
def mod():
@named_sequence("__transform_main", [any_op_t()], [])
def basic(variant_op: any_op_t()):
matmul = structured_match(
any_op_t(), variant_op, ops=["linalg.batch_reduce_matmul"]
)
top_func = get_parent_op(pdl.op_t(), matmul, op_name="func.func")
@apply_patterns(top_func)
def pats():
apply_patterns_canonicalization()
top_func = structured_match(any_op_t(), variant_op, ops=["func.func"])
apply_cse(top_func)
print(module_)
pm = PassManager.parse("builtin.module(transform-interpreter)")
pm.run(module_.operation)
# CHECK-LABEL: func.func @batch_reduce_matmul(
# CHECK-SAME: %[[VAL_0:.*]]: tensor<1x3x5xf32>, %[[VAL_1:.*]]: tensor<1x5x3xf32>, %[[VAL_2:.*]]: tensor<3x3xf32>) -> tensor<3x3xf32> {
# CHECK: %[[VAL_3:.*]] = linalg.batch_reduce_matmul ins(%[[VAL_0]], %[[VAL_1]] : tensor<1x3x5xf32>, tensor<1x5x3xf32>) outs(%[[VAL_2]] : tensor<3x3xf32>) -> tensor<3x3xf32>
# CHECK: return %[[VAL_3]] : tensor<3x3xf32>
# CHECK: }
print(module_)
|
