Skip to content

Commit 6e98c8c

Browse files
banach-spacebanach-space
authored
[mlir][linalg] Move vectorization tests for Tensor Ops (nfc) (#140877)
This patch reorganises vectorisation tests for tensor ops: * Tests for `tensor.pad` and `tensor.insert_slice` are extracted into dedicated files under a new `vectorization/` subdirectory. * Test files for `tensor.extract` are renamed and moved to the same subdirectory. Goals: * Unify test file naming. * Better organise the growing set of tests, which are currently hard to navigate. This is also a preparatory step for upcoming changes. I’ll soon be updating the vectorisation logic for `tensor.pad` and `tensor.insert_slice`. With the new structure in place, follow-up changes will be easier to review: * Only tests related to those ops will be updated. * Changes (e.g., to masking logic) will be isolated to the relevant tests. This patch implements part of #141025 - please see the ticket for full context.
1 parent 16fdb4f commit 6e98c8c

File tree

8 files changed

+599
-593
lines changed

8 files changed

+599
-593
lines changed

‎mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir‎

Lines changed: 0 additions & 315 deletions
Large diffs are not rendered by default.

‎mlir/test/Dialect/Linalg/vectorization.mlir‎

Lines changed: 0 additions & 277 deletions
Original file line numberDiff line numberDiff line change
@@ -580,133 +580,6 @@ module attributes {transform.with_named_sequence} {
580580
}
581581
}
582582

583-
// -----
584-
585-
// CHECK-LABEL: func @test_masked_vectorize_pad
586-
func.func @test_masked_vectorize_pad(
587-
%0 : tensor<?x?xf32>, %h0 : index, %h1 : index)
588-
-> tensor<2x4xf32>
589-
{
590-
// CHECK-DAG: %[[c42:.*]] = arith.constant 4.243000e+01 : f32
591-
// CHECK-DAG: %[[c0:.*]] = arith.constant 0 : index
592-
// CHECK-DAG: %[[c0_0:.*]] = arith.constant 0 : index
593-
// CHECK: %[[d0:.*]] = tensor.dim {{.*}} : tensor<?x?xf32>
594-
// CHECK: %[[d1:.*]] = tensor.dim {{.*}} : tensor<?x?xf32>
595-
// CHECK: %[[mask:.*]] = vector.create_mask %[[d0]], %[[d1]] : vector<2x4xi1>
596-
// CHECK: %[[masked_read:.*]] = vector.mask %[[mask]] {
597-
// CHECK-SAME: vector.transfer_read %{{.*}}[%[[c0_0]], %[[c0_0]]], %[[c42]]
598-
// CHECK-SAME: {in_bounds = [true, true]} : tensor<?x?xf32>, vector<2x4xf32>
599-
// CHECK-SAME: } : vector<2x4xi1> -> vector<2x4xf32>
600-
// CHECK-DAG: %[[c0_1:.*]] = arith.constant 0 : index
601-
// CHECK-DAG: %[[empty:.*]] = tensor.empty() : tensor<2x4xf32>
602-
// CHECK: vector.transfer_write %[[masked_read]], %[[empty]][%[[c0_1]], %[[c0_1]]]
603-
// CHECK-SAME: {in_bounds = [true, true]} : vector<2x4xf32>, tensor<2x4xf32>
604-
%cst = arith.constant 42.43 : f32
605-
%c0 = arith.constant 0 : index
606-
%1 = tensor.pad %0 low[0, %c0] high[%h0, %h1] {
607-
^bb0(%hh1: index, %hh2: index):
608-
tensor.yield %cst : f32
609-
} : tensor<?x?xf32> to tensor<2x4xf32>
610-
return %1: tensor<2x4xf32>
611-
}
612-
613-
module attributes {transform.with_named_sequence} {
614-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
615-
%0 = transform.structured.match ops{["tensor.pad"]} in %arg1
616-
: (!transform.any_op) -> !transform.any_op
617-
transform.structured.vectorize %0 vector_sizes [2, 4] : !transform.any_op
618-
transform.yield
619-
}
620-
}
621-
622-
// -----
623-
624-
// CHECK: #[[MAP:.+]] = affine_map<()[s0, s1] -> (s0 + s1)>
625-
// CHECK: func @test_masked_vectorize_dynamic_pad
626-
func.func @test_masked_vectorize_dynamic_pad(
627-
%0 : tensor<?x?xf32>, %h0 : index, %h1 : index)
628-
-> tensor<?x?xf32>
629-
{
630-
// CHECK-DAG: %[[c42:.*]] = arith.constant 4.243000e+01 : f32
631-
// CHECK-DAG: %[[c0:.*]] = arith.constant 0 : index
632-
// CHECK-DAG: %[[res_d0:.+]] = affine.apply #[[MAP]]()
633-
// CHECK-DAG: %[[res_d1:.+]] = affine.apply #[[MAP]]()
634-
// CHECK: %[[c0_2:.*]] = arith.constant 0 : index
635-
// CHECK: %[[d0:.*]] = tensor.dim {{.*}} : tensor<?x?xf32>
636-
// CHECK: %[[d1:.*]] = tensor.dim {{.*}} : tensor<?x?xf32>
637-
// CHECK: %[[mask:.*]] = vector.create_mask %[[d0]], %[[d1]] : vector<2x4xi1>
638-
// CHECK: %[[masked_read:.*]] = vector.mask %[[mask]] {
639-
// CHECK-SAME: vector.transfer_read %{{.*}}[%[[c0_2]], %[[c0_2]]], %[[c42]]
640-
// CHECK-SAME: {in_bounds = [true, true]} : tensor<?x?xf32>, vector<2x4xf32>
641-
// CHECK-SAME: } : vector<2x4xi1> -> vector<2x4xf32>
642-
// CHECK-DAG: %[[empty:.*]] = tensor.empty(%[[res_d0]], %[[res_d1]]) : tensor<?x?xf32>
643-
// CHECK-DAG: %[[c0_3:.*]] = arith.constant 0 : index
644-
// CHECK-DAG: %[[d2:.*]] = tensor.dim %[[empty]], {{.*}} : tensor<?x?xf32>
645-
// CHECK-DAG: %[[d3:.*]] = tensor.dim %[[empty]], {{.*}} : tensor<?x?xf32>
646-
// CHECK: %[[mask_2:.*]] = vector.create_mask %[[d2]], %[[d3]] : vector<2x4xi1>
647-
// CHECK: %[[masked_write:.*]] = vector.mask %[[mask_2]] {
648-
// CHECK-SAME: vector.transfer_write %[[masked_read]], %[[empty]][%[[c0_3]], %[[c0_3]]]
649-
// CHECK-SAME: {in_bounds = [true, true]} : vector<2x4xf32>, tensor<?x?xf32>
650-
// CHECK: return %[[masked_write]] : tensor<?x?xf32>
651-
%cst = arith.constant 42.43 : f32
652-
%c0 = arith.constant 0 : index
653-
%1 = tensor.pad %0 low[0, %c0] high[%h0, %h1] {
654-
^bb0(%hh1: index, %hh2: index):
655-
tensor.yield %cst : f32
656-
} : tensor<?x?xf32> to tensor<?x?xf32>
657-
return %1: tensor<?x?xf32>
658-
}
659-
660-
module attributes {transform.with_named_sequence} {
661-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
662-
%0 = transform.structured.match ops{["tensor.pad"]} in %arg1
663-
: (!transform.any_op) -> !transform.any_op
664-
transform.structured.vectorize %0 vector_sizes [2, 4] : !transform.any_op
665-
transform.yield
666-
}
667-
}
668-
669-
// -----
670-
// This case is supported because low padding `%l0` is applied on
671-
// a unit dimension which is supported, non unit result dimension low
672-
// padding is currently unsupported.
673-
// CHECK-LABEL: func @test_masked_vectorize_non_zero_low_pad_unit_res_dim
674-
func.func @test_masked_vectorize_non_zero_low_pad_unit_res_dim(
675-
%0 : tensor<?x?xf32>, %h0 : index, %h1 : index, %l0 : index)
676-
-> tensor<1x4xf32>
677-
{
678-
// CHECK-DAG: %[[C42:.*]] = arith.constant 4.243000e+01 : f32
679-
// CHECK-DAG: %[[C0:.*]] = arith.constant 0 : index
680-
// CHECK: %[[C0_1:.*]] = arith.constant 0 : index
681-
// CHECK-DAG: %[[D0:.*]] = tensor.dim {{.*}} : tensor<?x?xf32>
682-
// CHECK-DAG: %[[D1:.*]] = tensor.dim {{.*}} : tensor<?x?xf32>
683-
// CHECK: %[[MASK:.*]] = vector.create_mask %[[D0]], %[[D1]] : vector<1x4xi1>
684-
// CHECK: %[[MASKED_READ:.*]] = vector.mask %[[MASK]] {
685-
// CHECK-SAME: vector.transfer_read %{{.*}}[%[[C0_1]], %[[C0_1]]], %[[C42]]
686-
// CHECK-SAME: {in_bounds = [true, true]} : tensor<?x?xf32>, vector<1x4xf32>
687-
// CHECK-SAME: } : vector<1x4xi1> -> vector<1x4xf32>
688-
// CHECK-DAG: %[[EMPTY:.*]] = tensor.empty() : tensor<1x4xf32>
689-
// CHECK-DAG: %[[C0_2:.*]] = arith.constant 0 : index
690-
// CHECK: %[[MASKED_WRITE:.*]] = vector.transfer_write %[[MASKED_READ]], %[[EMPTY]][%[[C0_2]], %[[C0_2]]]
691-
// CHECK-SAME: {in_bounds = [true, true]} : vector<1x4xf32>, tensor<1x4xf32>
692-
// CHECK: return %[[MASKED_WRITE]] : tensor<1x4xf32>
693-
%cst = arith.constant 42.43 : f32
694-
%c0 = arith.constant 0 : index
695-
%1 = tensor.pad %0 low[%l0, %c0] high[%h0, %h1] {
696-
^bb0(%hh1: index, %hh2: index):
697-
tensor.yield %cst : f32
698-
} : tensor<?x?xf32> to tensor<1x4xf32>
699-
return %1: tensor<1x4xf32>
700-
}
701-
702-
module attributes {transform.with_named_sequence} {
703-
transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
704-
%0 = transform.structured.match ops{["tensor.pad"]} in %arg1
705-
: (!transform.any_op) -> !transform.any_op
706-
transform.structured.vectorize %0 vector_sizes [1, 4] : !transform.any_op
707-
transform.yield
708-
}
709-
}
710583

711584
// -----
712585

@@ -1155,153 +1028,3 @@ func.func @test_vectorize_unpack_no_vector_sizes_permute(%source: tensor<4x7x4xf
11551028
}
11561029
}
11571030

1158-
// -----
1159-
1160-
///----------------------------------------------------------------------------------------
1161-
/// tensor.insert_slice
1162-
///----------------------------------------------------------------------------------------
1163-
1164-
func.func private @insert_slice_static_sizes(%source: tensor<?x3x?x1xi32>) -> tensor<5x3xi32> {
1165-
%c2 = arith.constant 2 : index
1166-
%init = tensor.empty() : tensor<5x3xi32>
1167-
1168-
%source_slice = tensor.extract_slice %source[0, %c2, 0, 0] [1, 1, 5, 1] [1, 1, 1, 1] : tensor<?x3x?x1xi32> to tensor<5x1xi32>
1169-
%res = tensor.insert_slice %source_slice into %init[0, %c2] [5, 1] [1, 1] : tensor<5x1xi32> into tensor<5x3xi32>
1170-
1171-
return %res : tensor<5x3xi32>
1172-
}
1173-
1174-
// CHECK-LABEL: func.func private @insert_slice_static_sizes(
1175-
// CHECK-SAME: %[[SEC:.*]]: tensor<?x3x?x1xi32>) -> tensor<5x3xi32> {
1176-
// CHECK: %[[C_2:.*]] = arith.constant 2 : index
1177-
// CHECK: %[[INIT:.*]] = tensor.empty() : tensor<5x3xi32>
1178-
// CHECK: %[[SRC_SLICE:.*]] = tensor.extract_slice %[[SEC]][0, %[[C_2]], 0, 0] [1, 1, 5, 1] [1, 1, 1, 1] : tensor<?x3x?x1xi32> to tensor<5x1xi32>
1179-
// CHECK-DAG: %[[PAD:.*]] = arith.constant 0 : i32
1180-
// CHECK-DAG: %[[C_5:.*]] = arith.constant 5 : index
1181-
// CHECK-DAG: %[[C_1:.*]] = arith.constant 1 : index
1182-
// CHECK: %[[MASK:.*]] = vector.create_mask %[[C_5]], %[[C_1]] : vector<8x1xi1>
1183-
// CHECK: %[[C0:.*]] = arith.constant 0 : index
1184-
// CHECK: %[[READ:.*]] = vector.mask %[[MASK]] { vector.transfer_read %[[SRC_SLICE]][%[[C0]], %[[C0]]], %[[PAD]] : tensor<5x1xi32>, vector<8x1xi32> } : vector<8x1xi1> -> vector<8x1xi32>
1185-
// CHECK: %[[C_0:.*]] = arith.constant 0 : index
1186-
// CHECK: %[[RES:.*]] = vector.mask %[[MASK]] { vector.transfer_write %[[READ]], %[[INIT]][%[[C_0]], %[[C_2]]] : vector<8x1xi32>, tensor<5x3xi32> } : vector<8x1xi1> -> tensor<5x3xi32>
1187-
// CHECK: return %[[RES]] : tensor<5x3xi32>
1188-
1189-
module attributes {transform.with_named_sequence} {
1190-
transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
1191-
%0 = transform.structured.match ops{["tensor.insert_slice"]} in %arg0 : (!transform.any_op) -> !transform.any_op
1192-
transform.structured.vectorize %0 vector_sizes [8, 1] : !transform.any_op
1193-
transform.yield
1194-
}
1195-
}
1196-
1197-
// -----
1198-
1199-
// One of the _source_ dimensions is dynamic (but _destination_ dimensions are static).
1200-
1201-
func.func private @insert_slice_dynamic_src_dim(%source: tensor<?x3x?x1xi32>, %size: index) -> tensor<5x3xi32> {
1202-
%c2 = arith.constant 2 : index
1203-
%init = tensor.empty() : tensor<5x3xi32>
1204-
1205-
%source_slice = tensor.extract_slice %source[0, %c2, 0, 0] [1, 1, %size, 1] [1, 1, 1, 1] : tensor<?x3x?x1xi32> to tensor<?x1xi32>
1206-
%res = tensor.insert_slice %source_slice into %init[0, %c2] [%size, 1] [1, 1] : tensor<?x1xi32> into tensor<5x3xi32>
1207-
1208-
return %res : tensor<5x3xi32>
1209-
}
1210-
1211-
// CHECK-LABEL: func.func private @insert_slice_dynamic_src_dim(
1212-
// CHECK-SAME: %[[SRC:.*]]: tensor<?x3x?x1xi32>,
1213-
// CHECK-SAME: %[[SIZE:.*]]: index) -> tensor<5x3xi32> {
1214-
// CHECK: %[[C_2:.*]] = arith.constant 2 : index
1215-
// CHECK: %[[INIT:.*]] = tensor.empty() : tensor<5x3xi32>
1216-
// CHECK: %[[SRC_SLICE:.*]] = tensor.extract_slice %[[SRC]][0, %[[C_2]], 0, 0] [1, 1, %[[SIZE]], 1] [1, 1, 1, 1] : tensor<?x3x?x1xi32> to tensor<?x1xi32>
1217-
// CHECK-DAG: %[[PAD:.*]] = arith.constant 0 : i32
1218-
// CHECK-DAG: %[[C_1:.*]] = arith.constant 1 : index
1219-
// CHECK: %[[MASK:.*]] = vector.create_mask %[[SIZE]], %[[C_1]] : vector<8x1xi1>
1220-
// CHECK: %[[C_0:.*]] = arith.constant 0 : index
1221-
// CHECK: %[[READ:.*]] = vector.mask %[[MASK]] { vector.transfer_read %[[SRC_SLICE]][%[[C_0]], %[[C_0]]], %[[PAD]] : tensor<?x1xi32>, vector<8x1xi32> } : vector<8x1xi1> -> vector<8x1xi32>
1222-
// CHECK: %[[C_0_1:.*]] = arith.constant 0 : index
1223-
// CHECK: %[[RES:.*]] = vector.mask %[[MASK]] { vector.transfer_write %[[READ]], %[[INIT]][%[[C_0_1]], %[[C_2]]] : vector<8x1xi32>, tensor<5x3xi32> } : vector<8x1xi1> -> tensor<5x3xi32>
1224-
// CHECK: return %[[RES]] : tensor<5x3xi32>
1225-
1226-
module attributes {transform.with_named_sequence} {
1227-
transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
1228-
%0 = transform.structured.match ops{["tensor.insert_slice"]} in %arg0 : (!transform.any_op) -> !transform.any_op
1229-
transform.structured.vectorize %0 vector_sizes [8, 1] : !transform.any_op
1230-
transform.yield
1231-
}
1232-
}
1233-
1234-
// -----
1235-
1236-
// One of the _destination_ dimensions is dynamic (but _source_ dimensions are static).
1237-
1238-
func.func private @insert_slice_dynamic_dest_dim(%source: tensor<?x3x?x1xi32>, %size: index) -> tensor<?x3xi32> {
1239-
%c2 = arith.constant 2 : index
1240-
%init = tensor.empty(%size) : tensor<?x3xi32>
1241-
1242-
%source_slice = tensor.extract_slice %source[0, %c2, 0, 0] [1, 1, 5, 1] [1, 1, 1, 1] : tensor<?x3x?x1xi32> to tensor<5x1xi32>
1243-
%res = tensor.insert_slice %source_slice into %init[0, %c2] [5, 1] [1, 1] : tensor<5x1xi32> into tensor<?x3xi32>
1244-
1245-
return %res : tensor<?x3xi32>
1246-
}
1247-
1248-
// CHECK-LABEL: func.func private @insert_slice_dynamic_dest_dim(
1249-
// CHECK-SAME: %[[SRC:.*]]: tensor<?x3x?x1xi32>,
1250-
// CHECK-SAME: %[[SIZE:.*]]: index) -> tensor<?x3xi32> {
1251-
// CHECK: %[[C_2:.*]] = arith.constant 2 : index
1252-
// CHECK: %[[INIT:.*]] = tensor.empty(%[[SIZE]]) : tensor<?x3xi32>
1253-
// CHECK: %[[SRC_SLICE:.*]] = tensor.extract_slice %[[SRC]][0, %[[C_2]], 0, 0] [1, 1, 5, 1] [1, 1, 1, 1] : tensor<?x3x?x1xi32> to tensor<5x1xi32>
1254-
// CHECK: %[[PAD:.*]] = arith.constant 0 : i32
1255-
// CHECK: %[[C_5:.*]] = arith.constant 5 : index
1256-
// CHECK: %[[C_1:.*]] = arith.constant 1 : index
1257-
// CHECK: %[[MASK:.*]] = vector.create_mask %[[C_5]], %[[C_1]] : vector<8x1xi1>
1258-
// CHECK: %[[C_0:.*]] = arith.constant 0 : index
1259-
// CHECK: %[[READ:.*]] = vector.mask %[[MASK]] { vector.transfer_read %[[SRC_SLICE]][%[[C_0]], %[[C_0]]], %[[PAD]] : tensor<5x1xi32>, vector<8x1xi32> } : vector<8x1xi1> -> vector<8x1xi32>
1260-
// CHECK: %[[C_0_1:.*]] = arith.constant 0 : index
1261-
// CHECK: %[[WRITE:.*]] = vector.mask %[[MASK]] { vector.transfer_write %[[READ]], %[[INIT]][%[[C_0_1]], %[[C_2]]] : vector<8x1xi32>, tensor<?x3xi32> } : vector<8x1xi1> -> tensor<?x3xi32>
1262-
// CHECK: return %[[WRITE]] : tensor<?x3xi32>
1263-
1264-
module attributes {transform.with_named_sequence} {
1265-
transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
1266-
%0 = transform.structured.match ops{["tensor.insert_slice"]} in %arg0 : (!transform.any_op) -> !transform.any_op
1267-
transform.structured.vectorize %0 vector_sizes [8, 1] : !transform.any_op
1268-
transform.yield
1269-
}
1270-
}
1271-
1272-
// -----
1273-
1274-
// At least one _source_ and one _destination_ dimensions are dynamic.
1275-
1276-
func.func private @insert_slice_dynamic_source_and_dest_dim(%source: tensor<?x3x?x1xi32>, %size: index) -> tensor<?x3xi32> {
1277-
%c2 = arith.constant 2 : index
1278-
%init = tensor.empty(%size) : tensor<?x3xi32>
1279-
1280-
%source_slice = tensor.extract_slice %source[0, %c2, 0, 0] [1, 1, %size, 1] [1, 1, 1, 1] : tensor<?x3x?x1xi32> to tensor<?x1xi32>
1281-
%res = tensor.insert_slice %source_slice into %init[0, %c2] [%size, 1] [1, 1] : tensor<?x1xi32> into tensor<?x3xi32>
1282-
1283-
return %res : tensor<?x3xi32>
1284-
}
1285-
1286-
// CHECK-LABEL: func.func private @insert_slice_dynamic_source_and_dest_dim(
1287-
// CHECK-SAME: %[[SRC:.*]]: tensor<?x3x?x1xi32>,
1288-
// CHECK-SAME: %[[SIZE:.*]]: index) -> tensor<?x3xi32> {
1289-
// CHECK: %[[C_2:.*]] = arith.constant 2 : index
1290-
// CHECK: %[[INIT:.*]] = tensor.empty(%[[SIZE]]) : tensor<?x3xi32>
1291-
// CHECK: %[[SRC_SIZE:.*]] = tensor.extract_slice %[[SRC]][0, %[[C_2]], 0, 0] [1, 1, %[[SIZE]], 1] [1, 1, 1, 1] : tensor<?x3x?x1xi32> to tensor<?x1xi32>
1292-
// CHECK: %[[PAD:.*]] = arith.constant 0 : i32
1293-
// CHECK: %[[C1:.*]] = arith.constant 1 : index
1294-
// CHECK: %[[MASK:.*]] = vector.create_mask %[[SIZE]], %[[C1]] : vector<8x1xi1>
1295-
// CHECK: %[[C0:.*]] = arith.constant 0 : index
1296-
// CHECK: %[[READ:.*]] = vector.mask %[[MASK]] { vector.transfer_read %[[SRC_SIZE]]{{\[}}%[[C0]], %[[C0]]], %[[PAD]] : tensor<?x1xi32>, vector<8x1xi32> } : vector<8x1xi1> -> vector<8x1xi32>
1297-
// CHECK: %[[C_0_1:.*]] = arith.constant 0 : index
1298-
// CHECK: %[[WRITE:.*]] = vector.mask %[[MASK]] { vector.transfer_write %[[READ]], %[[INIT]]{{\[}}%[[C_0_1]], %[[C_2]]] : vector<8x1xi32>, tensor<?x3xi32> } : vector<8x1xi1> -> tensor<?x3xi32>
1299-
// CHECK: return %[[WRITE]] : tensor<?x3xi32>
1300-
1301-
module attributes {transform.with_named_sequence} {
1302-
transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
1303-
%0 = transform.structured.match ops{["tensor.insert_slice"]} in %arg0 : (!transform.any_op) -> !transform.any_op
1304-
transform.structured.vectorize %0 vector_sizes [8, 1] : !transform.any_op
1305-
transform.yield
1306-
}
1307-
}

‎mlir/test/Dialect/Linalg/vectorize-tensor-extract.mlir‎ renamed to ‎mlir/test/Dialect/Linalg/vectorization/extract-with-patterns.mlir‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
// RUN: mlir-opt -split-input-file \
2-
// RUN: -transform-preload-library='transform-library-paths=%p/td/vectorize-with-patterns.mlir' \
2+
// RUN: -transform-preload-library='transform-library-paths=%p/../td/vectorize-with-patterns.mlir' \
33
// RUN: -transform-interpreter=entry-point=vectorize_with_patterns %s | FileCheck %s
44

55
//===----------------------------------------------------------------------===//

‎mlir/test/Dialect/Linalg/vectorize-tensor-extract-masked.mlir‎ renamed to ‎mlir/test/Dialect/Linalg/vectorization/extract.mlir‎

File renamed without changes.

0 commit comments

Comments
 (0)