"""A model with two matmul ops."""

      @def_function.function
      def matmul(self, input_tensor: core.Tensor) -> Mapping[str, core.Tensor]:
        """Performs a matrix multiplication.

        Args:
          input_tensor: Input tensor to matmul with the filter.

        Returns:
          A 'output' -> output tensor mapping
        """

                            const bool t_x, const bool t_y,
                            std::function<Output(Output, Output)> mul_fn) {
    TF_ASSERT_OK(root_.status());
    // Generate random (but compatible) shapes for matrix multiplication.
    std::vector<TensorShape> shapes;
    RandMatMulShapes(is_x_batch, is_y_batch, t_x, t_y, &shapes);
    TensorShape x_shape = shapes[0];
    TensorShape y_shape = shapes[1];
    TensorShape z_shape = shapes[2];

	if k > n {
		return z.SetInt64(0)
	}
	// reduce the number of multiplications by reducing k
	if k > n-k {
		k = n - k // C(n, k) == C(n, n-k)
	}
	// C(n, k) == n * (n-1) * ... * (n-k+1) / k * (k-1) * ... * 1
	//         == n * (n-1) * ... * (n-k+1) / 1 * (1+1) * ... * k
	//
	// Using the multiplicative formula produces smaller values
	// at each step, requiring fewer allocations and computations:
	//

        return self.has_bias() and self.bias_size != self.filters.shape[-1]

      @def_function.function
      def matmul(self, input_tensor: core.Tensor) -> Mapping[str, core.Tensor]:
        """Performs a matrix multiplication.

        Depending on self.has_bias and self.activation_fn, it may add a bias
        term or
        go through the activaction function.

        Args:

(RISBGZ x {r}) && r == s390x.NewRotateParams(32, 63, 0) => (MOVWZreg x)

// Use sign/zero extend instead of ANDW.
(ANDWconst [0x00ff] x) => (MOVBZreg x)
(ANDWconst [0xffff] x) => (MOVHZreg x)

// Strength reduce multiplication to the sum (or difference) of two powers of two.
//
// Examples:
//     5x -> 4x + 1x
//    10x -> 8x + 2x
//   120x -> 128x - 8x
//  -120x -> 8x - 128x
//

// Merge negation into fused multiply-add and multiply-subtract.
//
// Key:
//
//   [+ -](x * y [+ -] z).
//    _ N         A S
//                D U
//                D B
//
// Note: multiplication commutativity handled by rule generator.
(F(MADD|NMADD|MSUB|NMSUB)S neg:(FNEGS x) y z) && neg.Uses == 1 => (F(NMSUB|MSUB|NMADD|MADD)S x y z)

      CPred<"TFL::VerifyMulOpShapeConstraints(llvm::cast<MulOp>($_op))">>,
    DeclareOpInterfaceMethods<TFL_ArithmeticCount>]> {
  let summary = "Multiplication operator";

  let description = [{
    Element-wise multiplication operation.
  }];

  let arguments = (
    ins TFL_TensorOf<[F32, I32, UI32, I64, QI8, QUI8, QI16, I16, Complex<F<32>>]>:$lhs,

	for _, a := range prodZZ {
		arg := a
		testFunZZ(t, "MulZZ", MulZZ, arg)

		arg = argZZ{a.z, a.y, a.x}
		testFunZZ(t, "MulZZ symmetric", MulZZ, arg)
	}
}

// mulBytes returns x*y via grade school multiplication. Both inputs
// and the result are assumed to be in big-endian representation (to
// match the semantics of Int.Bytes and Int.SetBytes).
func mulBytes(x, y []byte) []byte {
	z := make([]byte, len(x)+len(y))

		// of the result.
		{
			name:         "POPCNT",
			argLength:    1,
			reg:          gp11,
			asm:          "POPCNT",
			typ:          "UInt64",
			clobberFlags: true,
		},

		// unsigned multiplication (64x64 → 128)
		//
		// Multiply the two 64-bit input operands together and place the 128-bit result into
		// an even-odd register pair. The second register in the target pair also contains

}

// This is only used for operations that may cause overflow.
var op2str2 = [...]string{
	token.ADD: "addition",
	token.SUB: "subtraction",
	token.XOR: "bitwise XOR",
	token.MUL: "multiplication",
	token.SHL: "shift",
}

// If typ is a type parameter, underIs returns the result of typ.underIs(f).
// Otherwise, underIs returns the result of f(under(typ)).
func underIs(typ Type, f func(Type) bool) bool {