}

// -----

func.func @testSlice(%arg0: tensor<2x3x5xf32>, %arg1: tensor<3xi32>, %arg2: tensor<3xi32>) -> tensor<?x3x5xf32> {
  %0 = "tfl.slice"(%arg0, %arg1, %arg2) : (tensor<2x3x5xf32>, tensor<3xi32>, tensor<3xi32>) -> tensor<?x3x5xf32>
  func.return %0 : tensor<?x3x5xf32>
}

// -----

func.func @testSliceBadBeginDimension(%arg0: tensor<2x3x5xf32>, %arg1: tensor<2xi32>, %arg2: tensor<3xi32>) -> tensor<?x3x5xf32> {

// CHECK:  return
}

func.func @select_v2_broadcast(%arg0: tensor<4xi1>, %arg1: tensor<3x4xf32>, %arg2: tensor<8x3x4xf32>) -> tensor<8x3x4xf32> {
  %0 = "tf.SelectV2"(%arg0, %arg1, %arg2) : (tensor<4xi1>, tensor<3x4xf32>, tensor<8x3x4xf32>) -> tensor<8x3x4xf32>
  func.return %0: tensor<8x3x4xf32>

// CHECK-LABEL: select_v2_broadcast
// CHECK:  "tfl.select_v2"(%arg0, %arg1, %arg2)
// CHECK:  return
}

func.func @testTranspose(tensor<2x3x4xf32>) -> tensor<3x2x4xf32> {
^bb0(%arg0: tensor<2x3x4xf32>):
  %cst = arith.constant dense<[2, 0, 1]> : tensor<3xi32>
  // expected-error @+1 {{requires y.shape[0] (3) to be equal to x.shape[perm[2]] (4)}}
  %0 = "tf.Transpose"(%arg0, %cst) {T = "tfdtype$DT_FLOAT", Tperm = "tfdtype$DT_INT32"} : (tensor<2x3x4xf32>, tensor<3xi32>) -> tensor<3x2x4xf32>
  func.return %0 : tensor<3x2x4xf32>
}

// -----

// Tests that a float `stablehlo.transpose` is not converted to `tfl.transpose`.

func.func @transpose_float(%arg0: tensor<2x3x4xf32>) -> tensor<4x3x2xf32> {
  %0 = stablehlo.transpose %arg0, dims = [2, 1, 0] : (tensor<2x3x4xf32>) -> tensor<4x3x2xf32>
  return %0 : tensor<4x3x2xf32>
}
// CHECK-LABEL: transpose_float
// CHECK-NOT: tfl.transpose
// CHECK: stablehlo.transpose

// -----

func.func private @dense_image_warp_invalid_input_shape(%arg0: tensor<2x4x4xf32>, %arg1: tensor<2x4x4x2xf32>) -> tensor<2x4x4x1xf32> attributes {tf._implements = "addons:DenseImageWarp"}

// expected-warning @+1 {{Flow should be a 4D float tensor}}
func.func private @dense_image_warp_invalid_flow_shape(%arg0: tensor<2x4x4x1xf32>, %arg1: tensor<2x4x4xf32>) -> tensor<2x4x4x1xf32> attributes {tf._implements = "addons:DenseImageWarp"}

  %arg0_shape = "tf.Const"() {value = dense<[1, 8, 4]> : tensor<3xi32>} : () -> tensor<3xi32>
  %arg0_reshaped = "tf.Reshape"(%arg0, %arg0_shape) : (tensor<8x4xf32>, tensor<3xi32>) -> tensor<1x8x4xf32>
  %zeroi2 = "tf.Const"() {value = dense<0> : tensor<2xi32>} : () -> tensor<2xi32>
  %axis = "tf.Const"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>

    return success();
  }
};

// Fuses Unpack with proceeding Concatenation to Reshape if output type has
// static shape and activation function is none. For example:
//
//   // %input: tensor<1x3x2xf32>
//   %unpack:3 = "tfl.unpack"(%input) {axis = 1 : i32, num = 3 : i32}
//   %res = "tfl.concatenation"(%unpack#0, %unpack#1, %unpack#2)
//        {axis = -1 : i32, fused_activation_function = "NONE"}
//

  func.func @shape_partial_eval(%arg0: tensor<1x?x7xf32>, %arg1: tensor<3x7xf32>) -> tensor<*xf32> {
    %0 = "tf.Shape"(%arg0) : (tensor<1x?x7xf32>) -> tensor<3xi32>

    // CHECK: tf.Reshape
    // CHECK-SAME: tensor<1x3x7xf32>
    %1 = "tf.Reshape"(%arg1, %0) : (tensor<3x7xf32>, tensor<3xi32>) -> tensor<*xf32>
    return %1 : tensor<*xf32>
  }

  // CHECK-LABEL: func @gather_concat_reshape