94 lines · plain
1// RUN: mlir-opt -test-linalg-pad-fusion -split-input-file %s | FileCheck %s2 3func.func @dynamic_pad_fusion(%arg0 : tensor<?x?xf32>, %arg1 : index, %arg2 : index,4 %arg3 : index, %arg4 : index, %arg5 : f32) -> tensor<?x?xf32> {5 %c0 = arith.constant 0 : index6 %c1 = arith.constant 1 : index7 %d0 = tensor.dim %arg0, %c0 : tensor<?x?xf32>8 %d1 = tensor.dim %arg0, %c1 : tensor<?x?xf32>9 %init = tensor.empty(%d0, %d1) : tensor<?x?xf32>10 %0 = linalg.generic {11 indexing_maps = [affine_map<(d0, d1) -> (d0, d1)>, affine_map<(d0, d1) -> (d0, d1)>],12 iterator_types = ["parallel", "parallel"]}13 ins(%arg0 : tensor<?x?xf32>) outs(%init : tensor<?x?xf32>) {14 ^bb0(%arg6 : f32, %arg7 : f32):15 %1 = arith.mulf %arg6, %arg6 : f3216 linalg.yield %1 : f3217 } -> tensor<?x?xf32>18 %1 = tensor.pad %0 low [%arg1, %arg2] high [%arg3, %arg4] {19 ^bb0(%arg6: index, %arg7 : index):20 tensor.yield %arg5 : f3221 } : tensor<?x?xf32> to tensor<?x?xf32>22 return %1 : tensor<?x?xf32>23}24 25// CHECK-DAG: #[[MAP:.+]] = affine_map<()[s0, s1, s2] -> (s0 + s1 + s2)>26// CHECK: func @dynamic_pad_fusion27// CHECK-SAME: %[[ARG0:.+]]: tensor<?x?xf32>28// CHECK-SAME: %[[ARG1:[a-zA-Z0-9]+]]: index29// CHECK-SAME: %[[ARG2:[a-zA-Z0-9]+]]: index30// CHECK-SAME: %[[ARG3:[a-zA-Z0-9]+]]: index31// CHECK-SAME: %[[ARG4:[a-zA-Z0-9]+]]: index32// CHECK-SAME: %[[ARG5:[a-zA-Z0-9]+]]: f3233// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index34// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index35// CHECK-DAG: %[[SOURCE:.+]] = linalg.generic36// CHECK-DAG: %[[SOURCE_D0:.+]] = tensor.dim %[[SOURCE]], %[[C0]]37// CHECK-DAG: %[[TARGET_D0:.+]] = affine.apply #[[MAP]]()[%[[SOURCE_D0]], %[[ARG1]], %[[ARG3]]]38// CHECK-DAG: %[[SOURCE_D1:.+]] = tensor.dim %[[SOURCE]], %[[C1]]39// CHECK-DAG: %[[TARGET_D1:.+]] = affine.apply #[[MAP]]()[%[[SOURCE_D1]], %[[ARG2]], %[[ARG4]]]40// CHECK: %[[INIT:.+]] = tensor.empty(%[[TARGET_D0]], %[[TARGET_D1]])41// CHECK: %[[FILL:.+]] = linalg.fill ins(%[[ARG5]]{{.*}}outs(%[[INIT]]42// CHECK-DAG: %[[SIZE_D0:.+]] = tensor.dim %[[SOURCE]], %[[C0]]43// CHECK-DAG: %[[SIZE_D1:.+]] = tensor.dim %[[SOURCE]], %[[C1]]44// CHECK: %[[SLICE:.+]] = tensor.extract_slice %[[FILL]]45// CHECK-SAME: [%[[ARG1]], %[[ARG2]]] [%[[SIZE_D0]], %[[SIZE_D1]]] [1, 1]46// CHECK: %[[SOURCE:.+]] = linalg.generic47// CHECK-SAME: outs(%[[SLICE]] : tensor<?x?xf32>)48// CHECK: %[[RESULT:.+]] = tensor.insert_slice %[[SOURCE]] into %[[FILL]]49// CHECK-SAME: [%[[ARG1]], %[[ARG2]]] [%[[SIZE_D0]], %[[SIZE_D1]]] [1, 1]50// CHECK: return %[[RESULT]]51 52// -----53 54func.func @mixed_pad_fusion(%arg0 : tensor<?x42xf32>, %arg1 : index, %arg2 : index,55 %arg3 : f32) -> tensor<49x?xf32> {56 %c0 = arith.constant 0 : index57 %d0 = tensor.dim %arg0, %c0 : tensor<?x42xf32>58 %init = tensor.empty(%d0) : tensor<42x?xf32>59 %0 = linalg.generic {60 indexing_maps = [affine_map<(d0, d1) -> (d0, d1)>, affine_map<(d0, d1) -> (d1, d0)>],61 iterator_types = ["parallel", "parallel"]}62 ins(%arg0 : tensor<?x42xf32>) outs(%init : tensor<42x?xf32>) {63 ^bb0(%arg4 : f32, %arg5 : f32):64 %1 = arith.mulf %arg4, %arg4 : f3265 linalg.yield %1 : f3266 } -> tensor<42x?xf32>67 %1 = tensor.pad %0 low [3, %arg1] high [4, %arg2] {68 ^bb0(%arg4: index, %arg5 : index):69 tensor.yield %arg3 : f3270 } : tensor<42x?xf32> to tensor<49x?xf32>71 return %1 : tensor<49x?xf32>72}73// CHECK-DAG: #[[MAP:.+]] = affine_map<()[s0, s1, s2] -> (s0 + s1 + s2)>74// CHECK: func @mixed_pad_fusion75// CHECK-SAME: %[[ARG0:.+]]: tensor<?x42xf32>76// CHECK-SAME: %[[ARG1:[a-zA-Z0-9]+]]: index77// CHECK-SAME: %[[ARG2:[a-zA-Z0-9]+]]: index78// CHECK-SAME: %[[ARG3:[a-zA-Z0-9]+]]: f3279// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index80// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index81// CHECK-DAG: %[[SOURCE:.+]] = linalg.generic82// CHECK-DAG: %[[SOURCE_D1:.+]] = tensor.dim %[[SOURCE]], %[[C1]]83// CHECK-DAG: %[[TARGET_D1:.+]] = affine.apply #[[MAP]]()[%[[SOURCE_D1]], %[[ARG1]], %[[ARG2]]]84// CHECK: %[[INIT:.+]] = tensor.empty(%[[TARGET_D1]])85// CHECK: %[[FILL:.+]] = linalg.fill ins(%[[ARG3]]{{.*}}outs(%[[INIT]]86// CHECK-DAG: %[[SIZE_D1:.+]] = tensor.dim %[[SOURCE]], %[[C1]]87// CHECK: %[[SLICE:.+]] = tensor.extract_slice %[[FILL]]88// CHECK-SAME: [3, %[[ARG1]]] [42, %[[SIZE_D1]]] [1, 1]89// CHECK: %[[SOURCE:.+]] = linalg.generic90// CHECK-SAME: outs(%[[SLICE]] : tensor<42x?xf32>)91// CHECK: %[[RESULT:.+]] = tensor.insert_slice %[[SOURCE]] into %[[FILL]]92// CHECK-SAME: [3, %[[ARG1]]] [42, %[[SIZE_D1]]] [1, 1]93// CHECK: return %[[RESULT]]94