diff options
Diffstat (limited to 'mlir/test/Dialect')
-rw-r--r-- | mlir/test/Dialect/Linalg/canonicalize.mlir | 55 | ||||
-rw-r--r-- | mlir/test/Dialect/Linalg/fusion-elementwise-ops.mlir | 40 | ||||
-rw-r--r-- | mlir/test/Dialect/Linalg/invalid.mlir | 10 | ||||
-rw-r--r-- | mlir/test/Dialect/Mesh/spmdization.mlir | 14 | ||||
-rw-r--r-- | mlir/test/Dialect/Tosa/ops.mlir | 14 | ||||
-rw-r--r-- | mlir/test/Dialect/Vector/vector-rewrite-narrow-types.mlir | 82 | ||||
-rw-r--r-- | mlir/test/Dialect/Vector/vector-transfer-flatten.mlir | 38 |
7 files changed, 147 insertions, 106 deletions
diff --git a/mlir/test/Dialect/Linalg/canonicalize.mlir b/mlir/test/Dialect/Linalg/canonicalize.mlir index 7adde31..206d7e9 100644 --- a/mlir/test/Dialect/Linalg/canonicalize.mlir +++ b/mlir/test/Dialect/Linalg/canonicalize.mlir @@ -102,17 +102,16 @@ func.func @tensor.cast.unranked(%a : tensor<*xf32>, %b : tensor<*xf32>, %c : ten // ----- // CHECK-LABEL: func @linalg_effects( -// CHECK-SAME: %[[A:[a-z0-9]*]]: tensor<?x?xf32> -// CHECK-SAME: %[[B:[a-z0-9]*]]: memref<?x?xf32> -// CHECK-SAME: %[[C:[a-z0-9]*]]: tensor<?x?xf32> -func.func @linalg_effects(%a : tensor<?x?xf32>, %b : memref<?x?xf32>, %c : tensor<?x?xf32>) { +func.func @linalg_effects( + %a : tensor<?x?xf32>, %b : tensor<?x?xf32>, %c : tensor<?x?xf32>, + %d : memref<?x?xf32>, %e : memref<?x?xf32>, %f : memref<?x?xf32>) { // CHECK-NOT: %{{.*}} = linalg.matmul - %t = linalg.matmul ins(%a, %b : tensor<?x?xf32>, memref<?x?xf32>) + %t = linalg.matmul ins(%a, %b : tensor<?x?xf32>, tensor<?x?xf32>) outs(%c : tensor<?x?xf32>) -> tensor<?x?xf32> // CHECK: linalg.matmul - linalg.matmul ins(%a, %c : tensor<?x?xf32>, tensor<?x?xf32>) - outs(%b : memref<?x?xf32>) + linalg.matmul ins(%d, %e : memref<?x?xf32>, memref<?x?xf32>) + outs(%f : memref<?x?xf32>) return } @@ -889,11 +888,11 @@ func.func @fold_multi_use_generic_op_with_consumer(%arg0 : tensor<?x?x?xf32>) -> // ----- #map = affine_map<(d0) -> (d0)> -func.func @identity_mixed(%arg0 : tensor<?xf32>, %arg1: memref<?xf32>) { +func.func @identity_buffer(%arg0 : memref<?xf32>, %arg1: memref<?xf32>) { linalg.generic { indexing_maps = [#map, #map], iterator_types = ["parallel"] - } ins(%arg0 : tensor<?xf32>) + } ins(%arg0 : memref<?xf32>) outs(%arg1 : memref<?xf32>) { ^bb0(%arg2 : f32, %arg3 : f32): linalg.yield %arg2 : f32 @@ -901,14 +900,13 @@ func.func @identity_mixed(%arg0 : tensor<?xf32>, %arg1: memref<?xf32>) { return } -// There was a crash in EraseIdentityGenericOp for generic with mixed semantics. -// For now, check generic remained unchanged. -// CHECK-LABEL: func @identity_mixed -// CHECK-SAME: (%[[ARG1:.*]]: tensor<?xf32>, %[[ARG2:.*]]: memref<?xf32>) +// Do not erase ops with buffer semantics. +// CHECK-LABEL: func @identity_buffer +// CHECK-SAME: (%[[ARG1:.*]]: memref<?xf32>, %[[ARG2:.*]]: memref<?xf32>) // CHECK: linalg.generic { // CHECK-SAME: indexing_maps = [#map, #map], // CHECK-SAME: iterator_types = ["parallel"] -// CHECK-SAME: } ins(%[[ARG1]] : tensor<?xf32>) +// CHECK-SAME: } ins(%[[ARG1]] : memref<?xf32>) // CHECK-SAME: outs(%[[ARG2]] : memref<?xf32>) { // ----- @@ -916,12 +914,12 @@ func.func @identity_mixed(%arg0 : tensor<?xf32>, %arg1: memref<?xf32>) { // Just make sure that we don't crash. // CHECK-LABEL: func @dedeplicate_regression_test -func.func @dedeplicate_regression_test(%0: tensor<4xf32>, %1: memref<4xf32>) { +func.func @dedeplicate_regression_test(%0: tensor<4xf32>, %1: tensor<4xf32>) { %36 = linalg.generic {indexing_maps = [affine_map<(d0) -> (d0)>, affine_map<(d0) -> (d0)>, affine_map<(d0) -> (d0)>], iterator_types = ["parallel"]} - ins(%1, %1 : memref<4xf32>, memref<4xf32>) + ins(%1, %1 : tensor<4xf32>, tensor<4xf32>) outs(%0 : tensor<4xf32>) { ^bb0(%in: f32, %in_24: f32, %out: f32): linalg.yield %in : f32 @@ -937,31 +935,6 @@ func.func @dedeplicate_regression_test(%0: tensor<4xf32>, %1: memref<4xf32>) { // ----- -#map = affine_map<(d0) -> (d0)> -func.func @cast_producer_mixed(%arg0 : tensor<5xf32>, %arg1: memref<?xf32>) { - %0 = tensor.cast %arg0 : tensor<5xf32> to tensor<?xf32> - linalg.generic { - indexing_maps = [#map, #map], - iterator_types = ["parallel"] - } ins(%0 : tensor<?xf32>) - outs(%arg1 : memref<?xf32>) { - ^bb0(%arg2 : f32, %arg3 : f32): - linalg.yield %arg2 : f32 - } - return -} - -// We need a mixed linalg as a bridge between tensor and memref worlds. -// CHECK-LABEL: func @cast_producer_mixed -// CHECK-SAME: (%[[ARG1:.*]]: tensor<5xf32>, %[[ARG2:.*]]: memref<?xf32>) -// CHECK: linalg.generic { -// CHECK-SAME: indexing_maps = [#map, #map], -// CHECK-SAME: iterator_types = ["parallel"] -// CHECK-SAME: } ins(%[[ARG1]] : tensor<5xf32>) -// CHECK-SAME: outs(%[[ARG2]] : memref<?xf32>) { - -// ----- - // CHECK-LABEL: dead_softmax func.func @dead_softmax(%arg0: tensor<16x64x256xf32>) -> tensor<16x64x256xf32> { %0 = tensor.empty() : tensor<16x64x256xf32> diff --git a/mlir/test/Dialect/Linalg/fusion-elementwise-ops.mlir b/mlir/test/Dialect/Linalg/fusion-elementwise-ops.mlir index 9d8421c..15a4f6c 100644 --- a/mlir/test/Dialect/Linalg/fusion-elementwise-ops.mlir +++ b/mlir/test/Dialect/Linalg/fusion-elementwise-ops.mlir @@ -1110,43 +1110,3 @@ module { // CHECK-DAG: %[[T3:.+]] = arith.addf %[[T2]], %[[B1]] // CHECK: linalg.yield %[[T3]] : f32 // CHECK: return %[[GENERIC]] - -// ----- - -// CHECK-DAG: [[$MAP0:#[a-zA-Z0-9_]*]] = affine_map<(d0, d1) -> (d0, d1)> -#map0 = affine_map<(d0, d1) -> (d0, d1)> - -// CHECK-LABEL: @mixed_fusion -func.func @mixed_fusion(%arg0: tensor<?x?xf32>, %arg1 : tensor<?x?xf32>, %arg2 : tensor<?x?xf32>, %arg8 : memref<?x?xf32>) -{ - %c0 = arith.constant 0 : index - %c1 = arith.constant 1 : index - %0 = tensor.dim %arg0, %c0 : tensor<?x?xf32> - %1 = tensor.dim %arg0, %c1 : tensor<?x?xf32> - %2 = tensor.empty(%0, %1) : tensor<?x?xf32> - %3 = linalg.generic {indexing_maps = [#map0, #map0, #map0], iterator_types = ["parallel", "parallel"]} - ins(%arg0, %arg1 : tensor<?x?xf32>, tensor<?x?xf32>) - outs(%2 : tensor<?x?xf32>) { - ^bb0(%arg3: f32, %arg4: f32, %arg5: f32): - %4 = arith.addf %arg3, %arg4 : f32 - linalg.yield %4 : f32 - } -> tensor<?x?xf32> - // CHECK: linalg.generic { - // CHECK-SAME: indexing_maps = {{\[}}[[$MAP0]], [[$MAP0]], [[$MAP0]], [[$MAP0]]{{\]}} - linalg.generic {indexing_maps = [#map0, #map0, #map0], iterator_types = ["parallel", "parallel"]} - ins(%3, %arg2 : tensor<?x?xf32>, tensor<?x?xf32>) - outs(%arg8 : memref<?x?xf32>) { - // CHECK: ^{{[a-zA-Z0-9_]*}} - // CHECK-SAME: [[ARG0:%[a-zA-Z0-9_]*]] - // CHECK-SAME: [[ARG1:%[a-zA-Z0-9_]*]] - // CHECK-SAME: [[ARG2:%[a-zA-Z0-9_]*]] - ^bb0(%arg5: f32, %arg6: f32, %arg7: f32): - // CHECK: [[T1:%[a-zA-Z0-9_]*]] = arith.addf [[ARG0]], [[ARG1]] - // CHECK-NOT: linalg.yield - // CHECK: arith.mulf [[T1]], [[ARG2]] - // CHECK: linalg.yield - %5 = arith.mulf %arg5, %arg6 : f32 - linalg.yield %5 : f32 - } - return -} diff --git a/mlir/test/Dialect/Linalg/invalid.mlir b/mlir/test/Dialect/Linalg/invalid.mlir index 916c04f..44c81c3 100644 --- a/mlir/test/Dialect/Linalg/invalid.mlir +++ b/mlir/test/Dialect/Linalg/invalid.mlir @@ -770,3 +770,13 @@ func.func @mmt4d_rank_mismatch(%A: tensor<16x16x8x1xf32>, -> tensor<8x8xf32> return %res : tensor<8x8xf32> } + +// ----- + +func.func @mixed_semantics(%a: tensor<?x?xf32>, %b: tensor<?x?xf32>, %c: memref<?x?xf32>) { + // expected-error @+1 {{expected to have pure tensor or buffer semantics}} + linalg.matmul ins(%a, %b: tensor<?x?xf32>, tensor<?x?xf32>) + outs(%c: memref<?x?xf32>) + return +} + diff --git a/mlir/test/Dialect/Mesh/spmdization.mlir b/mlir/test/Dialect/Mesh/spmdization.mlir index 2fb8029..572d3eb 100644 --- a/mlir/test/Dialect/Mesh/spmdization.mlir +++ b/mlir/test/Dialect/Mesh/spmdization.mlir @@ -127,3 +127,17 @@ func.func @multiple_chained_ops( // CHECK: return %[[RESHARD3]] : tensor<1xi8> return %7 : tensor<2xi8> } + +// CHECK-LABEL: func @incomplete_sharding +func.func @incomplete_sharding( + // CHECK-SAME: %[[ARG:.*]]: tensor<4x16xf32> + %arg0: tensor<8x16xf32> +// CHECK-SAME: -> tensor<4x16xf32> { +) -> tensor<8x16xf32> { + %0 = mesh.shard %arg0 to <@mesh_1d, [[0]]> annotate_for_users : tensor<8x16xf32> + // CHECK: %[[RES:.*]] = tosa.sigmoid %[[ARG]] : (tensor<4x16xf32>) -> tensor<4x16xf32> + %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32> + %2 = mesh.shard %1 to <@mesh_1d, [[0]]> : tensor<8x16xf32> + // CHECK: return %[[RES]] : tensor<4x16xf32> + return %2 : tensor<8x16xf32> +} diff --git a/mlir/test/Dialect/Tosa/ops.mlir b/mlir/test/Dialect/Tosa/ops.mlir index 3d68464..01b2707 100644 --- a/mlir/test/Dialect/Tosa/ops.mlir +++ b/mlir/test/Dialect/Tosa/ops.mlir @@ -376,6 +376,13 @@ func.func @test_clz(%arg0: tensor<13x21x3xi32>) -> tensor<13x21x3xi32> { } // ----- +// CHECK-LABEL: cos +func.func @test_cos(%arg0: tensor<13x21x3xf32>) -> tensor<13x21x3xf32> { + %0 = tosa.cos %arg0 : (tensor<13x21x3xf32>) -> tensor<13x21x3xf32> + return %0 : tensor<13x21x3xf32> +} + +// ----- // CHECK-LABEL: exp func.func @test_exp(%arg0: tensor<13x21x3xf32>) -> tensor<13x21x3xf32> { %0 = tosa.exp %arg0 : (tensor<13x21x3xf32>) -> tensor<13x21x3xf32> @@ -425,6 +432,13 @@ func.func @test_rsqrt(%arg0: tensor<13x21x3xf32>) -> tensor<13x21x3xf32> { } // ----- +// CHECK-LABEL: sin +func.func @test_sin(%arg0: tensor<13x21x3xf32>) -> tensor<13x21x3xf32> { + %0 = tosa.sin %arg0 : (tensor<13x21x3xf32>) -> tensor<13x21x3xf32> + return %0 : tensor<13x21x3xf32> +} + +// ----- // CHECK-LABEL: select func.func @test_select(%arg0: tensor<1x1x1xi1>, %arg1: tensor<13x21x3xf32>, %arg2: tensor<13x21x3xf32>) -> tensor<13x21x3xf32> { %0 = tosa.select %arg0, %arg1, %arg2 : (tensor<1x1x1xi1>, tensor<13x21x3xf32>, tensor<13x21x3xf32>) -> tensor<13x21x3xf32> diff --git a/mlir/test/Dialect/Vector/vector-rewrite-narrow-types.mlir b/mlir/test/Dialect/Vector/vector-rewrite-narrow-types.mlir index 02063a8..94e78ce 100644 --- a/mlir/test/Dialect/Vector/vector-rewrite-narrow-types.mlir +++ b/mlir/test/Dialect/Vector/vector-rewrite-narrow-types.mlir @@ -195,53 +195,89 @@ func.func @f3ext(%a: vector<5xi8>) -> vector<8xi17> { // CHECK-LABEL: func.func @aligned_extsi( func.func @aligned_extsi(%a: vector<8xi4>) -> vector<8xi32> { - // CHECK: arith.shli - // CHECK: arith.shrsi - // CHECK: arith.shrsi - // CHECK: vector.shuffle - // CHECK: arith.extsi %{{.*}} : vector<8xi8> to vector<8xi32> +// CHECK-SAME: %[[IN:.*]]: vector<8xi4>) -> vector<8xi32> { +// CHECK: %[[I4_BITS:.*]] = arith.constant dense<4> : vector<4xi8> +// CHECK: %[[BITCAST:.*]] = vector.bitcast %[[IN]] : vector<8xi4> to vector<4xi8> +// CHECK: %[[SHL_LOW:.*]] = arith.shli %[[BITCAST]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[LOW:.*]] = arith.shrsi %[[SHL_LOW]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[HIGH:.*]] = arith.shrsi %[[BITCAST]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[INTERLEAVE:.*]] = vector.interleave %[[LOW]], %[[HIGH]] : vector<4xi8> +// CHECK: %[[I32:.*]] = arith.extsi %[[INTERLEAVE]] : vector<8xi8> to vector<8xi32> %0 = arith.extsi %a : vector<8xi4> to vector<8xi32> return %0 : vector<8xi32> } +// CHECK-LABEL: func.func @aligned_extsi_2d( +func.func @aligned_extsi_2d(%a: vector<8x32xi4>) -> vector<8x32xi32> { +// CHECK-SAME: %[[IN:.*]]: vector<8x32xi4>) -> vector<8x32xi32> { +// CHECK: %[[I4_BITS:.*]] = arith.constant dense<4> : vector<8x16xi8> +// CHECK: %[[BITCAST:.*]] = vector.bitcast %[[IN]] : vector<8x32xi4> to vector<8x16xi8> +// CHECK: %[[SHL_LOW:.*]] = arith.shli %[[BITCAST]], %[[I4_BITS]] : vector<8x16xi8> +// CHECK: %[[LOW:.*]] = arith.shrsi %[[SHL_LOW]], %[[I4_BITS]] : vector<8x16xi8> +// CHECK: %[[HIGH:.*]] = arith.shrsi %[[BITCAST]], %[[I4_BITS]] : vector<8x16xi8> +// CHECK: %[[INTERLEAVE:.*]] = vector.interleave %[[LOW]], %[[HIGH]] : vector<8x16xi8> +// CHECK: %[[I32:.*]] = arith.extsi %[[INTERLEAVE]] : vector<8x32xi8> to vector<8x32xi32> + %0 = arith.extsi %a : vector<8x32xi4> to vector<8x32xi32> + return %0 : vector<8x32xi32> +} + // CHECK-LABEL: func.func @aligned_extsi_base_case( func.func @aligned_extsi_base_case(%a: vector<8xi4>) -> vector<8xi8> { - // CHECK: arith.shli - // CHECK: arith.shrsi - // CHECK: arith.shrsi - // CHECK: vector.shuffle - // CHECK-NOT: arith.extsi +// CHECK-SAME: %[[IN:.*]]: vector<8xi4>) -> vector<8xi8> { +// CHECK: %[[I4_BITS:.*]] = arith.constant dense<4> : vector<4xi8> +// CHECK: %[[BITCAST:.*]] = vector.bitcast %[[IN]] : vector<8xi4> to vector<4xi8> +// CHECK: %[[SHL_LOW:.*]] = arith.shli %[[BITCAST]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[LOW:.*]] = arith.shrsi %[[SHL_LOW]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[HIGH:.*]] = arith.shrsi %[[BITCAST]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[INTERLEAVE:.*]] = vector.interleave %[[LOW]], %[[HIGH]] : vector<4xi8> %0 = arith.extsi %a : vector<8xi4> to vector<8xi8> return %0 : vector<8xi8> } // CHECK-LABEL: func.func @aligned_sitofp( func.func @aligned_sitofp(%a: vector<8xi4>) -> vector<8xf32> { - // CHECK: arith.shli - // CHECK: arith.shrsi - // CHECK: arith.shrsi - // CHECK: shuffle - // CHECK: arith.sitofp %{{.*}} : vector<8xi8> to vector<8xf32> +// CHECK-SAME: %[[IN:.*]]: vector<8xi4>) -> vector<8xf32> { +// CHECK: %[[I4_BITS:.*]] = arith.constant dense<4> : vector<4xi8> +// CHECK: %[[BITCAST:.*]] = vector.bitcast %[[IN]] : vector<8xi4> to vector<4xi8> +// CHECK: %[[SHL_LOW:.*]] = arith.shli %[[BITCAST]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[LOW:.*]] = arith.shrsi %[[SHL_LOW]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[HIGH:.*]] = arith.shrsi %[[BITCAST]], %[[I4_BITS]] : vector<4xi8> +// CHECK: %[[INTERLEAVE:.*]] = vector.interleave %[[LOW]], %[[HIGH]] : vector<4xi8> +// CHECK: %[[F32:.*]] = arith.sitofp %[[INTERLEAVE]] : vector<8xi8> to vector<8xf32> %0 = arith.sitofp %a : vector<8xi4> to vector<8xf32> return %0 : vector<8xf32> } +// CHECK-LABEL: func.func @aligned_sitofp_2d( +func.func @aligned_sitofp_2d(%a: vector<8x32xi4>) -> vector<8x32xf32> { +// CHECK-SAME: %[[IN:.*]]: vector<8x32xi4>) -> vector<8x32xf32> { +// CHECK: %[[I4_BITS:.*]] = arith.constant dense<4> : vector<8x16xi8> +// CHECK: %[[BITCAST:.*]] = vector.bitcast %[[IN]] : vector<8x32xi4> to vector<8x16xi8> +// CHECK: %[[SHL_LOW:.*]] = arith.shli %[[BITCAST]], %[[I4_BITS]] : vector<8x16xi8> +// CHECK: %[[LOW:.*]] = arith.shrsi %[[SHL_LOW]], %[[I4_BITS]] : vector<8x16xi8> +// CHECK: %[[HIGH:.*]] = arith.shrsi %[[BITCAST]], %[[I4_BITS]] : vector<8x16xi8> +// CHECK: %[[INTERLEAVE:.*]] = vector.interleave %[[LOW]], %[[HIGH]] : vector<8x16xi8> +// CHECK: %[[F32:.*]] = arith.sitofp %[[INTERLEAVE]] : vector<8x32xi8> to vector<8x32xf32> + %0 = arith.sitofp %a : vector<8x32xi4> to vector<8x32xf32> + return %0 : vector<8x32xf32> +} + // CHECK-LABEL: func.func @i4_transpose( -// CHECK-SAME: %[[A:[0-9a-z]*]] func.func @i4_transpose(%a: vector<8x16xi4>) -> vector<16x8xi4> { - // CHECK: %[[EXT:.*]] = arith.extsi %[[A]] : vector<8x16xi4> to vector<8x16xi8> - // CHECK: %[[TRANS:.*]] = vector.transpose %[[EXT]], [1, 0] : vector<8x16xi8> to vector<16x8xi8> - // CHECK: %[[TRUNC:.*]] = arith.trunci %[[TRANS]] : vector<16x8xi8> to vector<16x8xi4> +// CHECK-SAME: %[[IN:.*]]: vector<8x16xi4>) -> vector<16x8xi4> { +// CHECK: %[[EXT:.*]] = vector.interleave +// CHECK: %[[TRANS:.*]] = vector.transpose %[[EXT]], [1, 0] : vector<8x16xi8> to vector<16x8xi8> +// CHECK: %[[TRUNC:.*]] = arith.trunci %[[TRANS]] : vector<16x8xi8> to vector<16x8xi4> %0 = vector.transpose %a, [1, 0] : vector<8x16xi4> to vector<16x8xi4> return %0 : vector<16x8xi4> } // CHECK-LABEL: func.func @i7_transpose( -// CHECK-SAME: %[[A:[0-9a-z]*]] func.func @i7_transpose(%a: vector<8x16xi7>) -> vector<16x8xi7> { - // CHECK: %[[EXT:.*]] = arith.extsi %[[A]] : vector<8x16xi7> to vector<8x16xi8> - // CHECK: %[[TRANS:.*]] = vector.transpose %[[EXT]], [1, 0] : vector<8x16xi8> to vector<16x8xi8> - // CHECK: %[[TRUNC:.*]] = arith.trunci %[[TRANS]] : vector<16x8xi8> to vector<16x8xi7> +// CHECK-SAME: %[[IN:.*]]: vector<8x16xi7>) -> vector<16x8xi7> { +// CHECK: %[[EXT:.*]] = arith.extsi %[[IN]] : vector<8x16xi7> to vector<8x16xi8> +// CHECK: %[[TRANS:.*]] = vector.transpose %[[EXT]], [1, 0] : vector<8x16xi8> to vector<16x8xi8> +// CHECK: %[[TRUNC:.*]] = arith.trunci %[[TRANS]] : vector<16x8xi8> to vector<16x8xi7> %0 = vector.transpose %a, [1, 0] : vector<8x16xi7> to vector<16x8xi7> return %0 : vector<16x8xi7> } diff --git a/mlir/test/Dialect/Vector/vector-transfer-flatten.mlir b/mlir/test/Dialect/Vector/vector-transfer-flatten.mlir index 1775b5f..788ae9a 100644 --- a/mlir/test/Dialect/Vector/vector-transfer-flatten.mlir +++ b/mlir/test/Dialect/Vector/vector-transfer-flatten.mlir @@ -83,7 +83,7 @@ func.func @transfer_read_dims_mismatch_non_zero_indices( return } -// CHECK: #[[$ATTR_0:.+]] = affine_map<()[s0, s1] -> (s0 * 4 + s1 * 43)> +// CHECK: #[[$ATTR_0:.+]] = affine_map<()[s0, s1] -> (s0 * 24 + s1 * 6)> // CHECK-LABEL: func.func @transfer_read_dims_mismatch_non_zero_indices( // CHECK-SAME: %[[IDX_1:.*]]: index, %[[IDX_2:.*]]: index, @@ -92,7 +92,7 @@ func.func @transfer_read_dims_mismatch_non_zero_indices( // CHECK: %[[C_0:.*]] = arith.constant 0 : i32 // CHECK: %[[C_0_IDX:.*]] = arith.constant 0 : index // CHECK: %[[COLLAPSED_IN:.*]] = memref.collapse_shape %[[M_IN]] {{\[}}[0], [1, 2, 3]] : memref<1x43x4x6xi32> into memref<1x1032xi32> -// CHECK: %[[COLLAPSED_IDX:.*]] = affine.apply #[[$ATTR_0]]()[%[[IDX_2]], %[[IDX_1]]] +// CHECK: %[[COLLAPSED_IDX:.*]] = affine.apply #[[$ATTR_0]]()[%[[IDX_1]], %[[IDX_2]]] // CHECK: %[[READ:.*]] = vector.transfer_read %[[COLLAPSED_IN]][%[[C_0_IDX]], %[[COLLAPSED_IDX]]], %[[C_0]] {in_bounds = [true]} : memref<1x1032xi32>, vector<12xi32> // CHECK: %[[COLLAPSED_OUT:.*]] = memref.collapse_shape %[[M_OUT]] {{\[}}[0, 1, 2]] : memref<1x2x6xi32> into memref<12xi32> // CHECK: vector.transfer_write %[[READ]], %[[COLLAPSED_OUT]][%[[C_0_IDX]]] {in_bounds = [true]} : vector<12xi32>, memref<12xi32> @@ -459,3 +459,37 @@ func.func @fold_unit_dims_entirely(%arg0 : vector<8xi32>, // CHECK-128B-LABEL: func @fold_unit_dims_entirely( // CHECK-128B-NOT: memref.collapse_shape + +// ----- + +func.func @regression_non_contiguous_dim_read(%subview : memref<1x3x3x2xf32, strided<[40, 10, 2, 1], offset: ?>>, + %idx0 : index, %idx1 : index) -> vector<2x2xf32> { + %c0 = arith.constant 0 : index + %cst_1 = arith.constant 0.000000e+00 : f32 + %8 = vector.transfer_read %subview[%c0, %idx0, %idx1, %c0], %cst_1 {in_bounds = [true, true]} : memref<1x3x3x2xf32, strided<[40, 10, 2, 1], offset: ?>>, vector<2x2xf32> + return %8 : vector<2x2xf32> +} + +// CHECK: #[[$MAP:.+]] = affine_map<()[s0] -> (s0 * 2)> +// CHECK-LABEL: func.func @regression_non_contiguous_dim_read( +// CHECK: %[[COLLAPSE:.+]] = memref.collapse_shape %{{.*}} {{\[}}[0], [1], [2, 3]] : memref<1x3x3x2xf32, strided<[40, 10, 2, 1], offset: ?>> into memref<1x3x6xf32, strided<[40, 10, 1], offset: ?>> +// CHECK: %[[APPLY:.*]] = affine.apply #[[$MAP]]() + +// CHECK-128B-LABEL: func @regression_non_contiguous_dim_read( +// CHECK-128B: memref.collapse_shape + +// ----- + +func.func @unsupported_non_contiguous_dim_write(%value : vector<2x2xf32>, + %subview : memref<1x3x3x2xf32, strided<[40, 10, 2, 1], offset: ?>>, + %idx0 : index, %idx1 : index) { + %c0 = arith.constant 0 : index + vector.transfer_write %value, %subview[%c0, %idx0, %idx1, %c0] {in_bounds = [true, true]} : vector<2x2xf32>, memref<1x3x3x2xf32, strided<[40, 10, 2, 1], offset: ?>> + return +} + +// CHECK-LABEL: func.func @unsupported_non_contiguous_dim_write( +// CHECK-NOT: memref.collapse_shape + +// CHECK-128B-LABEL: func @unsupported_non_contiguous_dim_write( +// CHECK-128B-NOT: memref.collapse_shape |