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 {

	// looks strange but the purpose is to arrive at an integer division
	// (e.g. if retainExtraPercent = 12.5, then we get a divisor of 8)
	// that also avoids the overflow from a multiplication.
	gcPercentGoal += gcPercentGoal / (1.0 / (retainExtraPercent / 100.0))
	// Align it to a physical page boundary to make the following calculations
	// a bit more exact.

}

// This is only used for operations that may cause overflow.
var op2str2 = [...]string{
	syntax.Add: "addition",
	syntax.Sub: "subtraction",
	syntax.Xor: "bitwise XOR",
	syntax.Mul: "multiplication",
	syntax.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 {

	var r int64
	if rate <= 0 {
		r = 0 // disable profiling
	} else if rate == 1 {
		r = 1 // profile everything
	} else {
		// convert ns to cycles, use float64 to prevent overflow during multiplication
		r = int64(float64(rate) * float64(ticksPerSecond()) / (1000 * 1000 * 1000))
		if r == 0 {
			r = 1
		}
	}

	atomic.Store64(&blockprofilerate, uint64(r))
}

	//
	// The constants are obscured in this way for efficiency. The denominator
	// of the fraction is always a power-of-two for a quick division, so that
	// the numerator is a single constant integer multiplication.
	triggerRatioDen = 64

	// The minimum trigger constant was chosen empirically: given a sufficiently
	// fast/scalable allocator with 48 Ps that could drive the trigger ratio

func (p *PriorityQueue) calculateBackoffDuration(podInfo *framework.QueuedPodInfo) time.Duration {
	duration := p.podInitialBackoffDuration
	for i := 1; i < podInfo.Attempts; i++ {
		// Use subtraction instead of addition or multiplication to avoid overflow.
		if duration > p.podMaxBackoffDuration-duration {
			return p.podMaxBackoffDuration
		}
		duration += duration
	}
	return duration
}

      input_signature=[
          tensor_spec.TensorSpec(shape=[1, 4], dtype=dtypes.float32)
      ]
  )
  def matmul(self, matmul_input: core.Tensor) -> Mapping[str, core.Tensor]:
    """Performs a matrix multiplication.

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

    Returns:
      A map of: output key -> output result.
    """
    out = math_ops.matmul(matmul_input, self.matmul_filters)