out.montgomeryMul(out, out, m)

			// Select x^k in constant time from the table.
			k := uint((b >> j) & 0b1111)
			for i := range table {
				tmp.assign(ctEq(k, uint(i+1)), table[i])
			}

			// Multiply by x^k, discarding the result if k = 0.
			tmp.montgomeryMul(out, tmp, m)
			out.assign(not(ctEq(k, 0)), tmp)
		}
	}

	return out.montgomeryReduction(m)
}

//
// Consider a sequence of op:
//
// ```
// node 1: %0 = stablehlo.constant
// node 2: %1 = stablehlo.constant
// node 3: %2 = stablehlo.add %0, %1
// node 4: %3 = stablehlo.multiply %2, %1
// node 5: return %3
// ```
//
// In Backward Liveliness analysis, the liveliness for each node above becomes:
// live_in[5] = use[5]   U (live_out[5] - def[5])

  %0 = "mhlo.broadcast_in_dim"(%arg0) <{broadcast_dimensions = dense<[0, 1]> : tensor<2xi64>}> : (tensor<1x1xf32>) -> tensor<1x1000xf32>
  %1 = mhlo.multiply %0, %arg1 : tensor<1x1000xf32>
  %2 = mhlo.multiply %arg1, %0 : tensor<1x1000xf32>
  func.return %1, %2 : tensor<1x1000xf32>, tensor<1x1000xf32>
}

// CHECK-LABEL:   func @broadcast_mul_chlo(

    checkBigIntegerConversion("127.255.255.255", BigInteger.valueOf(Integer.MAX_VALUE));
    checkBigIntegerConversion(
        "255.255.255.254", BigInteger.valueOf(Integer.MAX_VALUE).multiply(BigInteger.valueOf(2)));
    checkBigIntegerConversion(
        "255.255.255.255", BigInteger.ONE.shiftLeft(32).subtract(BigInteger.ONE));
  }

  public void testFromIpv6BigIntegerValid() {

	//
	//   x = xh*b + x0  (0 <= x0 < b)
	//   y = yh*b + y0  (0 <= y0 < b)
	//   b = 1<<(_W*k)  ("base" of digits xi, yi)
	//
	k := karatsubaLen(n, karatsubaThreshold)
	// k <= n

	// multiply x0 and y0 via Karatsuba
	x0 := x[0:k]              // x0 is not normalized
	y0 := y[0:k]              // y0 is not normalized
	z = z.make(max(6*k, m+n)) // enough space for karatsuba of x0*y0 and full result of x*y

			}
			// Order of operations matters.
			// For very fast benchmarks, prevIters ~= prevns.
			// If you divide first, you get 0 or 1,
			// which can hide an order of magnitude in execution time.
			// So multiply first, then divide.
			n = goalns * prevIters / prevns
			// Run more iterations than we think we'll need (1.2x).
			n += n / 5
			// Don't grow too fast in case we had timing errors previously.
			n = min(n, 100*last)

* **Deep Learning**: this is a sub-field of Machine Learning, so, the same applies. It's just that there is not a single spreadsheet of numbers to multiply, but a huge set of them, and in many cases, you use a special processor to build and / or use those models.

### Concurrency + Parallelism: Web + Machine Learning

		p.firstProg = prog
	} else {
		p.lastProg.Link = prog
	}
	p.lastProg = prog
	if doLabel {
		p.pc++
		for _, label := range p.pendingLabels {
			if p.labels[label] != nil {
				p.errorf("label %q multiply defined", label)
				return
			}
			p.labels[label] = prog
		}
		p.pendingLabels = p.pendingLabels[0:0]
	}
	prog.Pc = p.pc
	if *flags.Debug {
		fmt.Println(p.lineNum, prog)
	}
	if testOut != nil {

    bool is_signed, bool narrow_range, bool legacy_float_scale = false,
    bool use_fake_quant_num_bits = false);

// Returns the quantized type of a bias input, given the quantized types of
// other operands which are multiply-accumulated (the bias is added to the
// accumulated value).
quant::QuantizedType GetUniformQuantizedTypeForBias(
    const std::vector<quant::QuantizedType>& op_types, int adjusted_quant_dim,

(Sub32F ...) => (FSUBS ...)
(Sub64F ...) => (FSUB ...)

(Min(32|64)F x y) && buildcfg.GOPPC64 >= 9 => (XSMINJDP x y)
(Max(32|64)F x y) && buildcfg.GOPPC64 >= 9 => (XSMAXJDP x y)

// Combine 64 bit integer multiply and adds
(ADD l:(MULLD x y) z) && buildcfg.GOPPC64 >= 9 && l.Uses == 1 && clobber(l) => (MADDLD x y z)

(Mod16 x y) => (Mod32 (SignExt16to32 x) (SignExt16to32 y))
(Mod16u x y) => (Mod32u (ZeroExt16to32 x) (ZeroExt16to32 y))