%0 = "tf.Einsum"(%arg0, %arg1) {T = "tfdtype$DT_FLOAT", equation = "ijk,ikm->ijm"}: (tensor<3x4x5xf32>, tensor<3x5x6xf32>) -> tensor<3x4x6xf32>
  func.return %0 : tensor<3x4x6xf32>
  // CHECK-LABEL: einsum_basic
  // CHECK: "tf.BatchMatMulV2"(%arg0, %arg1) <{adj_x = false, adj_y = false}> : (tensor<3x4x5xf32>, tensor<3x5x6xf32>) -> tensor<3x4x6xf32>
}

  // CHECK: %[[LHS_SPLIT:.*]]:6 = "tf.Split"(%[[SPLITTING_AXIS]], %[[LHS_RESHAPED]]) : (tensor<i32>, tensor<6x5x4xf32>) -> (tensor<1x5x4xf32>, tensor<1x5x4xf32>, tensor<1x5x4xf32>, tensor<1x5x4xf32>, tensor<1x5x4xf32>, tensor<1x5x4xf32>)
  // CHECK: %[[LHS_1:.*]] = "tf.Reshape"(%[[LHS_SPLIT]]#0, %[[MATMUL_LHS_SHAPE]]) : (tensor<1x5x4xf32>, tensor<2xi64>) -> tensor<5x4xf32>

  // CHECK:      %[[SLICE3:.*]] = "tf.Slice"(%[[CONCAT]], %[[CST5]], %[[CST6]]) : (tensor<3x8x4xi32>, tensor<3xi64>, tensor<3xi64>) -> tensor<3x6x4xi32>
  // CHECK:      %[[CONCAT1:.*]] = "tf.ConcatV2"(%[[SLICE2]], %[[SLICE3]], %[[CST7]]) : (tensor<3x2x4xi32>, tensor<3x6x4xi32>, tensor<i32>) -> tensor<3x8x4xi32>
  // CHECK:      return %[[CONCAT1]] : tensor<3x8x4xi32>
}

// CHECK-NEXT:    %[[RESHAPED_BMM:.*]] = mhlo.reshape %[[BMM_0]]
// CHECK-NEXT:    return %[[RESHAPED_BMM]] : tensor<3x5x1x4xf32>
}


func.func @convert_dot_general_repeated(%arg0: tensor<1x1x1024xf32>, %arg1: tensor<1024x1024xf32>) -> tensor<1x1x1024xf32> {

func.func @batchmatmulv2_basic(%arg0: tensor<1x4x2xf32>, %arg1: tensor<3x2x4xf32>) -> tensor<3x4x4xf32> {
// CHECK-LABEL:   func @batchmatmulv2_basic
// CHECK-SAME:        ([[LHS:%.*]]: tensor<1x4x2xf32>, [[RHS:%.*]]: tensor<3x2x4xf32>) -> tensor<3x4x4xf32>
// CHECK:           [[LHSSHAPE:%.*]] = shape.shape_of [[LHS]] : tensor<1x4x2xf32>
// CHECK:           [[RHSSHAPE:%.*]] = shape.shape_of [[RHS]] : tensor<3x2x4xf32>
// CHECK:           [[CM2:%.*]] = arith.constant -2 : index

  // CHECK-LABEL: test_batch_matmul_to_einsum
  // CHECK: "tf.Einsum"(%arg0, %arg1) <{equation = "...mk,...kn->...mn"}> : (tensor<1x2x3xf32>, tensor<3x4xf32>) -> tensor<1x2x4xf32>
  %0 = "tf.BatchMatMul"(%arg0, %arg1) {adj_x = false, adj_y = false} : (tensor<1x2x3xf32>, tensor<3x4xf32>) -> tensor<1x2x4xf32>
  func.return %0: tensor<1x2x4xf32>
}

  %0 = "tf.BatchMatMulV2"(%arg0, %arg1) {T = f32, adj_x = false, adj_y = false, grad_x = false, grad_y = false, device = ""} : (tensor<1x4x2xf32>, tensor<3x2x4xf32>) -> tensor<3x4x4xf32>
  func.return %0 : tensor<3x4x4xf32>
}

      return %0 : tensor<1x1x4xf32>
    }
  )mlir";

  OwningOpRef<ModuleOp> module_op =
      ParseModuleOpString(kModuleXlaCallModuleOpWithDefaultQuantizationMethod);
  ASSERT_TRUE(module_op);

    func.func @main() -> (tensor<1x4x4xf32>) {
      %%arg0 = "tf.Const"() {value = dense<-3.0> : tensor<1x4x2xf32>} : () -> tensor<1x4x2xf32>
      %%arg1 = "tf.Const"() {value = dense<-3.0> : tensor<1x2x4xf32>} : () -> tensor<1x2x4xf32>

      %%1 = "tf.%s"(%%arg0, %%arg1) {T = f32, adj_x = false, adj_y = false, grad_x = false, grad_y = false, device = ""} : (tensor<1x4x2xf32>, tensor<1x2x4xf32>) -> tensor<1x4x4xf32>

func.func @broadcast_mul_implicit_no_fold(%arg0: tensor<5x7xf32>, %arg1: tensor<5xf32>) -> tensor<3x5x7xf32> {
  %cst = arith.constant dense<[3, 5, 7]> : tensor<3xi32>
  %0 = "tf.BroadcastTo"(%arg1, %cst) : (tensor<5xf32>, tensor<3xi32>) -> tensor<3x5x7xf32>
  %1 = "tf.Mul"(%arg0, %0) : (tensor<5x7xf32>, tensor<3x5x7xf32>) -> tensor<3x5x7xf32>
  func.return %1 : tensor<3x5x7xf32>
  // CHECK: %[[C0:.*]] = arith.constant dense<[3, 5, 7]> : tensor<3xi32>