// RUN: mlir-opt --test-transform-dialect-interpreter %s | FileCheck %s

transform.with_pdl_patterns {
^bb0(%arg0: !pdl.operation):
  sequence %arg0 {
    ^bb0(%arg1: !pdl.operation):
      %0 = pdl_match @pdl_target in %arg1
      %1, %loops:3 = transform.structured.tile %0 {sizes = [4, 4, 4]}
  }

  pdl.pattern @pdl_target : benefit(1) {
    %args = operands
    %results = types
    %0 = operation "linalg.matmul"(%args : !pdl.range<value>) -> (%results : !pdl.range<type>)
    rewrite %0 with "transform.dialect"
  }
}

// CHECK-LABEL: func @tile_linalg_matmul(
// CHECK-SAME:    %[[TA:[0-9a-z]+]]: tensor<128x128xf32>
// CHECK-SAME:    %[[TB:[0-9a-z]+]]: tensor<128x128xf32>
// CHECK-SAME:    %[[TC:[0-9a-z]+]]: tensor<128x128xf32>
// CHECK-SAME:  -> tensor<128x128xf32> {
func.func @tile_linalg_matmul(
  %arg0: tensor<128x128xf32>, %arg1: tensor<128x128xf32>, %arg2: tensor<128x128xf32>)
    -> tensor<128x128xf32> {
//      CHECK: %[[TD0:.*]] = scf.for {{.*}} to {{.*}} step {{.*}} iter_args(%[[TC0:.*]] = %[[TC]]) -> (tensor<128x128xf32>) {
//      CHECK:   %[[TD1:.*]] = scf.for {{.*}} to {{.*}} step {{.*}} iter_args(%[[TC1:.*]] = %[[TC0]]) -> (tensor<128x128xf32>) {
//      CHECK:     %[[TD2:.*]] = scf.for {{.*}} to {{.*}} step {{.*}} iter_args(%[[TC2:.*]] = %[[TC1]]) -> (tensor<128x128xf32>) {
//      CHECK:       %[[sTA:.*]] = tensor.extract_slice %[[TA]][{{.*}}] : tensor<128x128xf32> to tensor<4x4xf32>
//      CHECK:       %[[sTB:.*]] = tensor.extract_slice %[[TB]][{{.*}}] : tensor<128x128xf32> to tensor<4x4xf32>
//      CHECK:       %[[sTC:.*]] = tensor.extract_slice %[[TC2]][{{.*}}] : tensor<128x128xf32> to tensor<4x4xf32>
//      CHECK:       %[[sTD:.*]] = linalg.matmul ins(%[[sTA]], %[[sTB]] : tensor<4x4xf32>, tensor<4x4xf32>)
// CHECK-SAME:                                   outs(%[[sTC]] : tensor<4x4xf32>)  -> tensor<4x4xf32>
//      CHECK:       %[[TD:.*]] = tensor.insert_slice %[[sTD]] into %[[TC2]][{{.*}}]  : tensor<4x4xf32> into tensor<128x128xf32>
//      CHECK:       scf.yield %[[TD]] : tensor<128x128xf32>
//      CHECK:     scf.yield %[[TD2]] : tensor<128x128xf32>
//      CHECK:   scf.yield %[[TD1]] : tensor<128x128xf32>
  %0 = linalg.matmul  ins(%arg0, %arg1: tensor<128x128xf32>, tensor<128x128xf32>)
                     outs(%arg2: tensor<128x128xf32>)
    -> tensor<128x128xf32>

//      CHECK: return %[[TD0]] : tensor<128x128xf32>
  return %0 : tensor<128x128xf32>
}