%outputs_16, %control_17 = tf_executor.island wraps "tf.Reshape"(%outputs_14, %outputs_6) {device = ""} : (tensor<16x16x16x?xf32>, tensor<2xi32>) -> tensor<?x16384xf32>
      %outputs_18, %control_19 = tf_executor.island wraps "tf.MatMul"(%outputs_16, %outputs_4) {device = "", transpose_a = false, transpose_b = false} : (tensor<?x16384xf32>, tensor<*xf32>) -> tensor<?x?xf32>

    Value input_rhs = bmm_op.getY();

    Value output_lhs =
        bmm_op.getAdjX() ? create_z_x_transpose_op(input_lhs) : input_lhs;

    // The rhs need to be transposed if adj_y == false AND this matmul will be
    // legalized to tfl.fully_connected
    Value output_rhs =
        !bmm_op.getAdjY() ? create_z_x_transpose_op(input_rhs) : input_rhs;

    Type output_type = bmm_op.getResult().getType();

def MakeOneDimValueBroadcastable : NativeCodeCall<
  "MakeOneDimValueBroadcastable($_builder, $_loc, $0, $1.getType().cast<ShapedType>())">;

// Match convolution op with "NHWC" data format or matmul op.
def SupportedAffineOpMatcher : NativeCodeCall<
  "MatchSupportedAffineOp($_self, $0, $1, $2)">;

// Checks if a value can be symetrically quantized.

	y.doinit()
	z.doinit()
	C.mpz_sub(&z.i[0], &x.i[0], &y.i[0])
	return z
}

// Mul sets z = x * y and returns z.
func (z *Int) Mul(x, y *Int) *Int {
	x.doinit()
	y.doinit()
	z.doinit()
	C.mpz_mul(&z.i[0], &x.i[0], &y.i[0])
	return z
}

// Div sets z = x / y, rounding toward zero, and returns z.
func (z *Int) Div(x, y *Int) *Int {
	x.doinit()
	y.doinit()
	z.doinit()