Skip to content

Commit 148f1c0

Browse files
banach-spacebanach-space
committed
[mlir][nfc] Update vectorize-tensor-extract.mlir (2/N)
Tests in "vectorize-tensor-extract.mlir" are inconsistent and would benefit from refactoring to: * Clearly categorize tests into "contiguous load," "gather load," and "scalar load + broadcast" cases, reflecting the structure of tensor.extract vectorization. * Unify variable naming (both MLIR and FileCheck). * Ensure all tests exercise unmasked vectorization (masked vectorization is covered in "vectorize-tensor-extract-masked.mlir"). * Improve and standardize formatting. These changes will make it easier to identify the test cases being exercised and simplify future maintenance or refactoring. This is patch 2/N in the series. Below is a summary of the changes in this patch. ---------------------------------------------------------------------- Summary of patch 2/N ---------------------------------------------------------------------- Makes all tests re-use the same TD sequence. This TD sequence has been extracted to a dedicted file: * "td/vectorize-with-patterns.mlir". ---------------------------------------------------------------------- **DEPENDS ON** * #118977 * #119079 Please only review the top commit
1 parent 39d7b84 commit 148f1c0

File tree

2 files changed

+13
-182
lines changed

2 files changed

+13
-182
lines changed

‎mlir/test/Dialect/Linalg/td/vectorize-with-patterns.mlir

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,10 @@
1+
module @transforms attributes { transform.with_named_sequence } {
2+
transform.named_sequence @vectorize_with_patterns(%module: !transform.any_op {transform.readonly}) {
3+
4+
%0 = transform.structured.match ops{["linalg.generic"]} in %module : (!transform.any_op) -> !transform.any_op
5+
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
6+
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
7+
8+
transform.yield
9+
}
10+
}

‎mlir/test/Dialect/Linalg/vectorize-tensor-extract.mlir

Lines changed: 3 additions & 182 deletions
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,6 @@
1-
// RUN: mlir-opt %s -transform-interpreter -split-input-file | FileCheck %s
1+
// RUN: mlir-opt -split-input-file \
2+
// RUN: -transform-preload-library='transform-library-paths=%p/td/vectorize-with-patterns.mlir' \
3+
// RUN: -transform-interpreter=entry-point=vectorize_with_patterns %s | FileCheck %s
24

35
#map0 = affine_map<(d0, d1, d2, d3) -> (d0, d2)>
46
#map1 = affine_map<(d0, d1, d2, d3) -> (d0, d1, d2, d3)>
@@ -27,15 +29,6 @@ func.func @vectorize_1d_tensor_extract(%arg0: tensor<3xf32>, %arg1: tensor<4x3xi
2729
// CHECK: %[[GATHER:.*]] = vector.gather %[[ARG0]][%[[C0]]] [%[[INDICES]]], %[[MASK]], %[[PASSTHRU]]
2830
// CHECK: vector.transfer_write %[[GATHER]]
2931

30-
module attributes {transform.with_named_sequence} {
31-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
32-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
33-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
34-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 : (!transform.any_op) -> !transform.any_op
35-
transform.yield
36-
}
37-
}
38-
3932
// -----
4033

4134
#map = affine_map<() -> ()>
@@ -59,15 +52,6 @@ func.func @extract_scalar_from_0d_into_0d(%src: tensor<f32>, %init: tensor<f32>)
5952
// CHECK: %[[READ:.*]] = vector.transfer_read %[[SRC]][], %[[PAD]] : tensor<f32>, vector<f32>
6053
// CHECK: vector.transfer_write %[[READ]], %[[INIT]][] : vector<f32>, tensor<f32>
6154

62-
module attributes {transform.with_named_sequence} {
63-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
64-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
65-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
66-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
67-
transform.yield
68-
}
69-
}
70-
7155
// -----
7256

7357
#map = affine_map<(n) -> (n)>
@@ -92,15 +76,6 @@ func.func @extract_scalar_from_0d_into_1d(%src: tensor<f32>, %init: tensor<1xf32
9276
// CHECK: %[[READ_BCAST:.*]] = vector.broadcast %[[READ]] : vector<f32> to vector<1xf32>
9377
// CHECK: vector.transfer_write %[[READ_BCAST]], %[[INIT]][%[[C0]]] {in_bounds = [true]} : vector<1xf32>, tensor<1xf32>
9478

95-
module attributes {transform.with_named_sequence} {
96-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
97-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
98-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
99-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
100-
transform.yield
101-
}
102-
}
103-
10479
// -----
10580

10681
#map = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
@@ -130,14 +105,6 @@ func.func @vectorize_nd_tensor_extract_scalar_broadcast(%arg0: tensor<3x3xf32>,
130105
// CHECK: %[[C0_2:.*]] = arith.constant 0 : index
131106
// CHECK: vector.transfer_write %[[READ]], %[[ARG_1]]{{\[}}%[[C0_2]], %[[C0_2]], %[[C0_2]]] : vector<1x1x3xf32>, tensor<1x1x3xf32>
132107

133-
module attributes {transform.with_named_sequence} {
134-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
135-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
136-
transform.structured.vectorize %0 { vectorize_nd_extract } : !transform.any_op
137-
transform.yield
138-
}
139-
}
140-
141108
// -----
142109

143110
#map = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
@@ -207,15 +174,6 @@ func.func @vectorize_nd_tensor_extract_transfer_read_basic_column(
207174
// CHECK: %[[RES:.*]] = vector.transfer_write %[[BCAST]], %[[OUTPUT]]{{\[}}%[[C0]], %[[C0]], %[[C0]]] {in_bounds = [true, true, true]} : vector<3x1x1xf32>, tensor<3x1x1xf32>
208175
// CHECK: return %[[RES]] : tensor<3x1x1xf32>
209176

210-
module attributes {transform.with_named_sequence} {
211-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
212-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
213-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
214-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
215-
transform.yield
216-
}
217-
}
218-
219177
// -----
220178

221179
func.func @vectorize_nd_tensor_extract_transfer_read_complex(%6: tensor<45x80x16xf32>, %arg0: index, %arg2: index, %arg1: index, %arg4: index, %extracted_slice : tensor<1x4xf32>) -> tensor<1x4xf32> {
@@ -259,15 +217,6 @@ func.func @vectorize_nd_tensor_extract_transfer_read_complex(%6: tensor<45x80x16
259217
// CHECK: return %[[VAL_21]] : tensor<1x4xf32>
260218
// CHECK: }
261219

262-
module attributes {transform.with_named_sequence} {
263-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
264-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
265-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
266-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
267-
transform.yield
268-
}
269-
}
270-
271220
// -----
272221

273222
#map0 = affine_map<(d0, d1, d2, d3) -> (d0, d2)>
@@ -309,15 +258,6 @@ func.func @vectorize_nd_tensor_extract_index_from_tensor(%arg0: tensor<3x3xf32>,
309258
// CHECK: %[[GATHER:.*]] = vector.gather %[[ARG0]][%[[C0]], %[[C0]]] [%[[T]]], %[[CST_1]], %[[PASSTHRU]] : tensor<3x3xf32>, vector<4x7x3x2xindex>, vector<4x7x3x2xi1>, vector<4x7x3x2xf32> into vector<4x7x3x2xf32>
310259
// CHECK: vector.transfer_write %[[GATHER]], %[[ARG4]][%[[C0]], %[[C0]], %[[C0]], %[[C0]]] {in_bounds = [true, true, true, true]} : vector<4x7x3x2xf32>, tensor<4x7x3x2xf32>
311260

312-
module attributes {transform.with_named_sequence} {
313-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
314-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
315-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
316-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
317-
transform.yield
318-
}
319-
}
320-
321261
// -----
322262

323263
#map = affine_map<(d0, d1) -> (d0, d1)>
@@ -339,15 +279,6 @@ func.func @vectorize_nd_tensor_extract_load_1d_column_vector_using_gather_load(%
339279
return %1 : tensor<8x1xf32>
340280
}
341281

342-
module attributes {transform.with_named_sequence} {
343-
transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
344-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg0 : (!transform.any_op) -> !transform.any_op
345-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
346-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 {vectorize_nd_extract} : (!transform.any_op) -> !transform.any_op
347-
transform.yield
348-
}
349-
}
350-
351282
// CHECK-LABEL: func.func @vectorize_nd_tensor_extract_load_1d_column_vector_using_gather_load
352283
// CHECK-SAME: %[[ARG0:.*]]: tensor<8x128x768xf32>
353284
// CHECK-SAME: %[[ARG1:.*]]: index
@@ -390,15 +321,6 @@ func.func @index_from_output_column_vector_gather_load(%src: tensor<8x128xf32>)
390321
return %res : tensor<8x1xf32>
391322
}
392323

393-
module attributes {transform.with_named_sequence} {
394-
transform.named_sequence @__transform_main(%arg2: !transform.any_op {transform.readonly}) {
395-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg2 : (!transform.any_op) -> !transform.any_op
396-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
397-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 {vectorize_nd_extract} : (!transform.any_op) -> !transform.any_op
398-
transform.yield
399-
}
400-
}
401-
402324
// CHECK-LABEL: func.func @index_from_output_column_vector_gather_load(
403325
// CHECK-SAME: %[[SRC:.*]]: tensor<8x128xf32>) -> tensor<8x1xf32> {
404326
// CHECK: %[[C128:.*]] = arith.constant dense<128> : vector<1x8xindex>
@@ -437,15 +359,6 @@ func.func @index_from_output_column_vector_contiguous_load(%src: tensor<8x128xf3
437359
return %res : tensor<8x1xf32>
438360
}
439361

440-
module attributes {transform.with_named_sequence} {
441-
transform.named_sequence @__transform_main(%arg2: !transform.any_op {transform.readonly}) {
442-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg2 : (!transform.any_op) -> !transform.any_op
443-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
444-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 {vectorize_nd_extract} : (!transform.any_op) -> !transform.any_op
445-
transform.yield
446-
}
447-
}
448-
449362
// CHECK-LABEL: func.func @index_from_output_column_vector_contiguous_load(
450363
// CHECK-SAME: %[[SRC:.*]]: tensor<8x128xf32>) -> tensor<8x1xf32> {
451364
// CHECK: %[[C0:.*]] = arith.constant 0 : index
@@ -497,15 +410,6 @@ func.func @vectorize_nd_tensor_extract_contiguous_and_gather(%arg0: tensor<6xf32
497410
// CHECK: %[[VAL_14:.*]] = vector.transfer_write %[[VAL_13]], %[[VAL_8]]{{\[}}%[[VAL_2]]] {in_bounds = [true]} : vector<5xf32>, tensor<5xf32>
498411
// CHECK: return %[[VAL_14]] : tensor<5xf32>
499412

500-
module attributes {transform.with_named_sequence} {
501-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
502-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
503-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
504-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
505-
transform.yield
506-
}
507-
}
508-
509413
// -----
510414

511415
// The vectorizer converts `affine.apply` so that the subsequent Ops can be vectorised based on the converted ops. Contiguous load.
@@ -540,15 +444,6 @@ func.func @vectorize_nd_tensor_extract_with_affine_apply_contiguous(%6: tensor<8
540444
// CHECK: return %[[VAL_12]] : tensor<1x4xf32>
541445
// CHECK: }
542446

543-
module attributes {transform.with_named_sequence} {
544-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
545-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
546-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
547-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
548-
transform.yield
549-
}
550-
}
551-
552447
// -----
553448

554449
func.func @vectorize_nd_tensor_extract_with_tensor_extract(%input_1: tensor<1x20xi32>, %input_2: tensor<257x24xf32>, %arg0 : index, %arg1 : index, %arg2 : index, %arg3 : index) -> tensor<1x1x4xf32> {
@@ -585,16 +480,6 @@ func.func @vectorize_nd_tensor_extract_with_tensor_extract(%input_1: tensor<1x20
585480
// for address calculation also satisfy the required conditions).
586481
// CHECK: vector.transfer_read %[[INPUT_2]][%{{.*}}, %{{.*}}, %{{.*}} {in_bounds = [true, true]} : tensor<257x24xf32>, vector<1x4xf32>
587482

588-
589-
module attributes {transform.with_named_sequence} {
590-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
591-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
592-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
593-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
594-
transform.yield
595-
}
596-
}
597-
598483
// -----
599484

600485
// The vectorizer converts `affine.apply` so that the subsequent Ops can be vectorised based on the converted ops. Gather load.
@@ -632,15 +517,6 @@ func.func @vectorize_nd_tensor_extract_with_affine_apply_gather(%6: tensor<80x16
632517
// CHECK: return %[[VAL_14]] : tensor<1x4xf32>
633518
// CHECK: }
634519

635-
module attributes {transform.with_named_sequence} {
636-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
637-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
638-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
639-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
640-
transform.yield
641-
}
642-
}
643-
644520
// -----
645521

646522
// Make sure that non-linear arithmetic operations (e.g. arith.maxsi) are allowed when calculating indices for load operations. Gather load.
@@ -674,15 +550,6 @@ func.func @vectorize_nd_tensor_extract_with_maxsi_gather(%arg0: tensor<80x16xf32
674550
// CHECK: return %[[VAL_10]] : tensor<1x4xf32>
675551
// CHECK: }
676552

677-
module attributes {transform.with_named_sequence} {
678-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
679-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
680-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
681-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
682-
transform.yield
683-
}
684-
}
685-
686553
// -----
687554

688555
// Make sure that non-linear arithmetic operations (e.g. arith.maxsi) are allowed when calculating indices for load operations. Contiguous load.
@@ -718,15 +585,6 @@ func.func @vectorize_nd_tensor_extract_with_maxsi_contiguous(%arg0: tensor<80x16
718585
// CHECK: return %[[VAL_9]] : tensor<1x4xf32>
719586
// CHECK: }
720587

721-
module attributes {transform.with_named_sequence} {
722-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
723-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
724-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
725-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
726-
transform.yield
727-
}
728-
}
729-
730588
// -----
731589

732590
// The vectorizer assumes it's a gather load whenever using a block argument to calculate an index.
@@ -759,15 +617,6 @@ func.func @vectorize_nd_tensor_extract_block_arg(%arg0: tensor<5x6xf32>, %arg1:
759617
// CHECK: return %[[VAL_12]] : tensor<5xf32>
760618
// CHECK: }
761619

762-
module attributes {transform.with_named_sequence} {
763-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
764-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
765-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
766-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
767-
transform.yield
768-
}
769-
}
770-
771620
// -----
772621

773622
#map1 = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
@@ -788,15 +637,6 @@ func.func @vectorize_0d_tensor_extract(%arg0: tensor<f32>, %arg2: tensor<1x1x3xf
788637
// CHECK: %[[EXTRACT:.*]] = vector.transfer_read %[[ARG_0]][], %{{.+}} : tensor<f32>
789638
// CHECK: vector.broadcast %[[EXTRACT]] : vector<f32> to vector<1x1x3xf32>
790639

791-
module attributes {transform.with_named_sequence} {
792-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
793-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
794-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
795-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
796-
transform.yield
797-
}
798-
}
799-
800640
// -----
801641

802642
#map = affine_map<(d0, d1, d2) -> (d0, d1, d2)>
@@ -833,16 +673,6 @@ func.func @vectorize_reverse_like_tensor_extract(%arg0: tensor<1x2x3xf32>, %arg1
833673
// CHECK: %[[GATHER:.*]] = vector.gather %[[ARG0]][%[[C0]], %[[C0]], %[[C0]]] [%[[T3]]], %[[MASK]], %[[PASSTHRU]]
834674
// CHECK: vector.transfer_write %[[GATHER]]
835675

836-
module attributes {transform.with_named_sequence} {
837-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
838-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
839-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
840-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
841-
transform.yield
842-
}
843-
}
844-
845-
846676
// -----
847677

848678
func.func @vectorize_scalar_read_with_broadcast_from_column_tensor(%init: tensor<1x1x4xi32>) -> tensor<1x1x4xi32> {
@@ -874,12 +704,3 @@ func.func @vectorize_scalar_read_with_broadcast_from_column_tensor(%init: tensor
874704
// CHECK: %[[READ:.*]] = vector.transfer_read %[[SRC]]{{\[}}%[[IDX_ELT]], %[[C0]]], %[[PAD]] : tensor<15x1xi32>, vector<i32>
875705
// CHECK: %[[READ_BCAST:.*]] = vector.broadcast %[[READ]] : vector<i32> to vector<1x1x4xi32>
876706
// CHECK: %[[RES:.*]] = vector.transfer_write %[[READ_BCAST]], %[[INIT]][%[[C0]], %[[C0]], %[[C0]]] {in_bounds = [true, true, true]} : vector<1x1x4xi32>, tensor<1x1x4xi32>
877-
878-
module attributes {transform.with_named_sequence} {
879-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
880-
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
881-
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
882-
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
883-
transform.yield
884-
}
885-
}

0 commit comments

Comments
 (0)