if (!fp_type) {
      return failure();
    }

    // result = (x - mean) * scale / sqrt(variance + epsilon) + offset
    // Let multiplier = scale / sqrt(variance + epsilon), to compute
    // (x - mean) * scale / sqrt(variance + epsilon) + offset,
    // is then to compute (x * multiplier) + (offset - mean * multiplier).

    auto epsilon = materializeEpsilon(

           const char* raw_device_name = nullptr);

// Computes square root of x element-wise.
Status Sqrt(AbstractContext* ctx, AbstractTensorHandle* const x,
            AbstractTensorHandle** y, const char* name = nullptr,
            const char* raw_device_name = nullptr);

// Computes the gradient for the sqrt of `x` wrt its input.
Status SqrtGrad(AbstractContext* ctx, AbstractTensorHandle* const y,

      <w>externref</w>
      <w>funcref</w>
      <w>jetbrains</w>
      <w>kotlinx</w>
      <w>ktor</w>
      <w>optref</w>
      <w>popcnt</w>
      <w>rotl</w>
      <w>rotr</w>
      <w>simd</w>
      <w>sqrt</w>
      <w>testsuite</w>
      <w>uninstantiable</w>
      <w>unintercepted</w>
      <w>unlinkable</w>
      <w>vtable</w>
      <w>wabt</w>
      <w>xopt</w>
    </words>
  </dictionary>

// operations. Specifically, performs the following calculation:
//
//   (x - mean) * scale / sqrt(variance + epsilon) + offset
//
// Let multiplier = scale / sqrt(variance + epsilon),
// to compute
//   (x - mean) * scale / sqrt(variance + epsilon) + offset,
// is then to compute
//   (x * multiplier) + (offset - mean * multiplier).
//

    // CHECK: [[LR_MULTIPLY:%.*]] = "tf.Mul"([[LR]], [[GRAD]]) : (tensor<f32>, tensor<f32>) -> tensor<f32>
    // CHECK: [[SQRT:%.*]] = "tf.Sqrt"([[NEW_ACC]]) : (tensor<*xf32>) -> tensor<*xf32>
    // CHECK: [[DIVISOR:%.*]] = "tf.AddV2"([[SQRT]], [[EPSILON]]) : (tensor<*xf32>, tensor<f32>) -> tensor<*xf32>

		// Check 1K total values, down from 64M.
		inc = 1 << 16
	}
	for i := 1; i < 1<<randomBitCount; i += inc {
		l, fl := math.Log2(float64(i)), runtime.Fastlog2(float64(i))
		d := l - fl
		e += d * d
	}
	e = math.Sqrt(e)

	if e > 1.0 {
		t.Fatalf("imprecision on fastlog2 implementation, want <=1.0, got %f", e)
	}

	return Vec2{v[0], v[0]}
}

func (v Vec2) B() Vec2 {
	return Vec2{1.0 / v.D(), 0}
}

func (v Vec2) C() Vec2 {
	return Vec2{v[0], v[0]}
}

func (v Vec2) D() float64 {
	return math.Sqrt(v[0])

 Const 5
 GreaterEqual 1
 MatMul 1
 Mul 5
 Select 2
 Sigmoid 3
 Snapshot 1
 Split 1
 Tanh 2
cluster 29 size 9
 Add 1
 Mul 2
 RealDiv 2
 Sqrt 2
 Sub 2
cluster 30 size 9
 Add 1
 Mul 2
 RealDiv 2
 Sqrt 2
 Sub 2
cluster 31 size 4
 Mul 3
 UnsortedSegmentSum 1
cluster 32 size 4
 Mul 3

}

func asin(x float64) float64 {
	if x == 0 {
		return x // special case
	}
	sign := false
	if x < 0 {
		x = -x
		sign = true
	}
	if x > 1 {
		return NaN() // special case
	}

	temp := Sqrt(1 - x*x)
	if x > 0.7 {
		temp = Pi/2 - satan(temp/x)
	} else {
		temp = satan(x / temp)
	}

	if sign {
		temp = -temp
	}
	return temp
}

// Acos returns the arccosine, in radians, of x.
//

	case IsInf(x, 0):
		if IsInf(x, -1) {
			return Pow(1/x, -y) // Pow(-0, -y)
		}
		switch {
		case y < 0:
			return 0
		case y > 0:
			return Inf(1)
		}
	case y == 0.5:
		return Sqrt(x)
	case y == -0.5:
		return 1 / Sqrt(x)
	}

	yi, yf := Modf(Abs(y))
	if yf != 0 && x < 0 {
		return NaN()
	}
	if yi >= 1<<63 {
		// yi is a large even int that will lead to overflow (or underflow to 0)