Skip to content

Commit 2adbf9e

Browse files
linuxlonelyeaglelinuxlonelyeagle
authored
[mlir][memref] Support test-compose-subview dynamic size (#146881)
Supports the case where the sizes of the subview op is dynamic.When there are more for loops in the tile algorithm, multiple subviews are performed and test-compose-subview does not work when the size operand of the subview ops is dynamic value.
1 parent d532d58 commit 2adbf9e

File tree

2 files changed

+48
-26
lines changed

2 files changed

+48
-26
lines changed

‎mlir/lib/Dialect/MemRef/Transforms/ComposeSubView.cpp‎

Lines changed: 0 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -80,10 +80,6 @@ struct ComposeSubViewOpPattern : public OpRewritePattern<memref::SubViewOp> {
8080
for (auto &&[opOffset, sourceOffset, sourceStride, opSize] :
8181
llvm::zip(op.getMixedOffsets(), sourceOp.getMixedOffsets(),
8282
sourceOp.getMixedStrides(), op.getMixedSizes())) {
83-
// We only support static sizes.
84-
if (isa<Value>(opSize)) {
85-
return failure();
86-
}
8783
sizes.push_back(opSize);
8884
Attribute opOffsetAttr = llvm::dyn_cast_if_present<Attribute>(opOffset),
8985
sourceOffsetAttr =

‎mlir/test/Transforms/compose-subview.mlir‎

Lines changed: 48 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,9 @@
11
// RUN: mlir-opt %s -test-compose-subview -split-input-file | FileCheck %s
22

33
// CHECK-LABEL: func.func @subview_strided(
4-
// CHECK-SAME: %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: 3456>> {
4+
// CHECK-SAME: %[[input:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: 3456>> {
55
func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: 3456>> {
6-
// CHECK: %[[VAL_1:.*]] = memref.subview %[[VAL_0]][3, 384] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: 3456>>
6+
// CHECK: {{.*}} = memref.subview %[[input]][3, 384] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: 3456>>
77
%0 = memref.subview %input[2, 256] [2, 256] [1, 1] : memref<4x1024xf32> to memref<2x256xf32, strided<[1024, 1], offset: 2304>>
88
%1 = memref.subview %0[1, 128] [1, 128] [1, 1] : memref<2x256xf32, strided<[1024, 1], offset: 2304>> to memref<1x128xf32, strided<[1024, 1], offset: 3456>>
99
return %1 : memref<1x128xf32, strided<[1024, 1], offset: 3456>>
@@ -12,9 +12,9 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, stri
1212
// -----
1313

1414
// CHECK-LABEL: func.func @subview_strided(
15-
// CHECK-SAME: %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x10xf32, strided<[1024, 1], offset: 3745>> {
15+
// CHECK-SAME: %[[input:.*]]: memref<4x1024xf32>) -> memref<1x10xf32, strided<[1024, 1], offset: 3745>> {
1616
func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x10xf32, strided<[1024, 1], offset: 3745>> {
17-
// CHECK: %[[VAL_1:.*]] = memref.subview %[[VAL_0]][3, 673] [1, 10] [1, 1] : memref<4x1024xf32> to memref<1x10xf32, strided<[1024, 1], offset: 3745>>
17+
// CHECK: {{.*}} = memref.subview %[[input]][3, 673] [1, 10] [1, 1] : memref<4x1024xf32> to memref<1x10xf32, strided<[1024, 1], offset: 3745>>
1818
%0 = memref.subview %input[1, 512] [3, 256] [1, 1] : memref<4x1024xf32> to memref<3x256xf32, strided<[1024, 1], offset: 1536>>
1919
%1 = memref.subview %0[1, 128] [2, 128] [1, 1] : memref<3x256xf32, strided<[1024, 1], offset: 1536>> to memref<2x128xf32, strided<[1024, 1], offset: 2688>>
2020
%2 = memref.subview %1[1, 33] [1, 10] [1, 1] : memref<2x128xf32, strided<[1024, 1], offset: 2688>> to memref<1x10xf32, strided<[1024, 1], offset: 3745>>
@@ -24,12 +24,12 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x10xf32, strid
2424
// -----
2525

2626
// CHECK-LABEL: func.func @subview_strided(
27-
// CHECK-SAME: %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
27+
// CHECK-SAME: %[[input:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
2828
func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
29-
// CHECK: %[[VAL_1:.*]] = arith.constant 3 : index
29+
// CHECK: %[[C3:.*]] = arith.constant 3 : index
3030
%cst_1 = arith.constant 1 : index
3131
%cst_2 = arith.constant 2 : index
32-
// CHECK: %[[VAL_2:.*]] = memref.subview %[[VAL_0]]{{\[}}%[[VAL_1]], 384] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
32+
// CHECK: {{.*}} = memref.subview %[[input]]{{\[}}%[[C3]], 384] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
3333
%0 = memref.subview %input[%cst_2, 256] [2, 256] [1, 1] : memref<4x1024xf32> to memref<2x256xf32, strided<[1024, 1], offset: ?>>
3434
%1 = memref.subview %0[%cst_1, 128] [1, 128] [1, 1] : memref<2x256xf32, strided<[1024, 1], offset: ?>> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
3535
return %1 : memref<1x128xf32, strided<[1024, 1], offset: ?>>
@@ -38,13 +38,13 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, stri
3838
// -----
3939

4040
// CHECK-LABEL: func.func @subview_strided(
41-
// CHECK-SAME: %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
41+
// CHECK-SAME: %[[input:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
4242
func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
43-
// CHECK: %[[VAL_1:.*]] = arith.constant 3 : index
43+
// CHECK: %[[C3:.*]] = arith.constant 3 : index
4444
%cst_2 = arith.constant 2 : index
45-
// CHECK: %[[VAL_2:.*]] = arith.constant 384 : index
45+
// CHECK: %[[C384:.*]] = arith.constant 384 : index
4646
%cst_128 = arith.constant 128 : index
47-
// CHECK: %[[VAL_3:.*]] = memref.subview %[[VAL_0]]{{\[}}%[[VAL_1]], %[[VAL_2]]] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
47+
// CHECK: {{.*}} = memref.subview %[[input]]{{\[}}%[[C3]], %[[C384]]] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
4848
%0 = memref.subview %input[%cst_2, 256] [2, 256] [1, 1] : memref<4x1024xf32> to memref<2x256xf32, strided<[1024, 1], offset: ?>>
4949
%1 = memref.subview %0[1, %cst_128] [1, 128] [1, 1] : memref<2x256xf32, strided<[1024, 1], offset: ?>> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
5050
return %1 : memref<1x128xf32, strided<[1024, 1], offset: ?>>
@@ -53,9 +53,9 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, stri
5353
// -----
5454

5555
// CHECK-LABEL: func.func @subview_strided(
56-
// CHECK-SAME: %[[VAL_0:.*]]: memref<8x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: 4480>> {
56+
// CHECK-SAME: %[[input:.*]]: memref<8x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: 4480>> {
5757
func.func @subview_strided(%input: memref<8x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: 4480>> {
58-
// CHECK: %[[VAL_1:.*]] = memref.subview %[[VAL_0]][4, 384] [1, 64] [4, 4] : memref<8x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: 4480>>
58+
// CHECK: {{.*}} = memref.subview %[[input]][4, 384] [1, 64] [4, 4] : memref<8x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: 4480>>
5959
%0 = memref.subview %input[2, 256] [2, 256] [2, 2] : memref<8x1024xf32> to memref<2x256xf32, strided<[2048, 2], offset: 2304>>
6060
%1 = memref.subview %0[1, 64] [1, 64] [2, 2] : memref<2x256xf32, strided<[2048, 2], offset: 2304>> to memref<1x64xf32, strided<[4096, 4], offset: 4480>>
6161
return %1 : memref<1x64xf32, strided<[4096, 4], offset: 4480>>
@@ -64,9 +64,9 @@ func.func @subview_strided(%input: memref<8x1024xf32>) -> memref<1x64xf32, strid
6464
// -----
6565

6666
// CHECK-LABEL: func.func @subview_strided(
67-
// CHECK-SAME: %[[VAL_0:.*]]: memref<30x30xf32>) -> memref<2x2xf32, strided<[240, 8], offset: 217>> {
67+
// CHECK-SAME: %[[input:.*]]: memref<30x30xf32>) -> memref<2x2xf32, strided<[240, 8], offset: 217>> {
6868
func.func @subview_strided(%input: memref<30x30xf32>) -> memref<2x2xf32, strided<[240, 8], offset: 217>> {
69-
// CHECK: %[[VAL_1:.*]] = memref.subview %[[VAL_0]][7, 7] [2, 2] [8, 8] : memref<30x30xf32> to memref<2x2xf32, strided<[240, 8], offset: 217>>
69+
// CHECK: {{.*}} = memref.subview %[[input]][7, 7] [2, 2] [8, 8] : memref<30x30xf32> to memref<2x2xf32, strided<[240, 8], offset: 217>>
7070
%0 = memref.subview %input[1, 1] [12, 12] [2, 2] : memref<30x30xf32> to memref<12x12xf32, strided<[60, 2], offset: 31>>
7171
%1 = memref.subview %0[1, 1] [5, 5] [2, 2] : memref<12x12xf32, strided<[60, 2], offset: 31>> to memref<5x5xf32, strided<[120, 4], offset: 93>>
7272
%2 = memref.subview %1[1, 1] [2, 2] [2, 2] : memref<5x5xf32, strided<[120, 4], offset: 93>> to memref<2x2xf32, strided<[240, 8], offset: 217>>
@@ -76,13 +76,13 @@ func.func @subview_strided(%input: memref<30x30xf32>) -> memref<2x2xf32, strided
7676
// -----
7777

7878
// CHECK-LABEL: func.func @subview_strided(
79-
// CHECK-SAME: %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
79+
// CHECK-SAME: %[[input:.*]]: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
8080
func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
81-
// CHECK: %[[VAL_1:.*]] = arith.constant 4 : index
81+
// CHECK: %[[C4:.*]] = arith.constant 4 : index
8282
%cst_2 = arith.constant 2 : index
83-
// CHECK: %[[VAL_2:.*]] = arith.constant 384 : index
83+
// CHECK: %[[C384:.*]] = arith.constant 384 : index
8484
%cst_64 = arith.constant 64 : index
85-
// CHECK: %[[VAL_3:.*]] = memref.subview %[[VAL_0]]{{\[}}%[[VAL_1]], %[[VAL_2]]] [1, 64] [4, 4] : memref<4x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
85+
// CHECK: {{.*}} = memref.subview %[[input]]{{\[}}%[[C4]], %[[C384]]] [1, 64] [4, 4] : memref<4x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
8686
%0 = memref.subview %input[%cst_2, 256] [2, 256] [2, 2] : memref<4x1024xf32> to memref<2x256xf32, strided<[2048, 2], offset: ?>>
8787
%1 = memref.subview %0[1, %cst_64] [1, 64] [2, 2] : memref<2x256xf32, strided<[2048, 2], offset: ?>> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
8888
return %1 : memref<1x64xf32, strided<[4096, 4], offset: ?>>
@@ -91,13 +91,39 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x64xf32, strid
9191
// -----
9292

9393
// CHECK-LABEL: func.func @subview_strided(
94-
// CHECK-SAME: %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
94+
// CHECK-SAME: %[[input:.*]]: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
9595
func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
96-
// CHECK: %[[VAL_1:.*]] = arith.constant 4 : index
96+
// CHECK: %[[C4:.*]] = arith.constant 4 : index
9797
%cst_1 = arith.constant 1 : index
9898
%cst_2 = arith.constant 2 : index
99-
// CHECK: %[[VAL_2:.*]] = memref.subview %[[VAL_0]]{{\[}}%[[VAL_1]], 384] [1, 64] [4, 4] : memref<4x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
99+
// CHECK: {{.*}} = memref.subview %[[input]]{{\[}}%[[C4]], 384] [1, 64] [4, 4] : memref<4x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
100100
%0 = memref.subview %input[%cst_2, 256] [2, 256] [2, 2] : memref<4x1024xf32> to memref<2x256xf32, strided<[2048, 2], offset: ?>>
101101
%1 = memref.subview %0[%cst_1, 64] [1, 64] [2, 2] : memref<2x256xf32, strided<[2048, 2], offset: ?>> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
102102
return %1 : memref<1x64xf32, strided<[4096, 4], offset: ?>>
103103
}
104+
105+
// -----
106+
107+
// CHECK-LABEL: func.func @single_dynamic_size_subview(
108+
// CHECK-SAME: %[[input:.*]]: memref<256x?xf32>,
109+
// CHECK-SAME: %{{.*}}: index,
110+
// CHECK-SAME: %[[SIZE_1:.*]]: index) -> memref<8x?xf32> {
111+
func.func @single_dynamic_size_subview(%input: memref<256x?xf32>, %size0 : index, %size1 : index) -> memref<8x?xf32>{
112+
%subview = memref.subview %input[0, 0][8, %size0][1, 1] : memref<256x?xf32> to memref<8x?xf32>
113+
%subview_1 = memref.subview %subview[0, 0][8, %size1][1, 1] : memref<8x?xf32> to memref<8x?xf32>
114+
// CHECK: %{{.*}} = memref.subview %[[input]][0, 0] [8, %[[SIZE_1]]] [1, 1] : memref<256x?xf32> to memref<8x?xf32>
115+
return %subview_1 : memref<8x?xf32>
116+
}
117+
118+
// -----
119+
120+
// CHECK-LABEL: func.func @all_dynamic_size_subview(
121+
// CHECK-SAME: %[[input:.*]]: memref<256x?xf32>,
122+
// CHECK-SAME: %{{.*}}: index,
123+
// CHECK-SAME: %[[SIZE1:.*]]: index) -> memref<?x?xf32> {
124+
func.func @all_dynamic_size_subview(%input: memref<256x?xf32>, %size0 : index, %size1 : index) -> memref<?x?xf32>{
125+
%subview = memref.subview %input[0, 0][%size0, %size0][1, 1] : memref<256x?xf32> to memref<?x?xf32>
126+
%subview_1 = memref.subview %subview[0, 0][%size1, %size1][1, 1] : memref<?x?xf32> to memref<?x?xf32>
127+
// CHECK: {{.*}} = memref.subview %[[input]][0, 0] {{\[}}%[[SIZE1]], %[[SIZE1]]] [1, 1] : memref<256x?xf32> to memref<?x?xf32>
128+
return %subview_1 : memref<?x?xf32>
129+
}

0 commit comments

Comments
 (0)