*/

const blockSize = 64 // A DCT block is 8x8.

type block [blockSize]int32

const (
	w1 = 2841 // 2048*sqrt(2)*cos(1*pi/16)
	w2 = 2676 // 2048*sqrt(2)*cos(2*pi/16)
	w3 = 2408 // 2048*sqrt(2)*cos(3*pi/16)
	w5 = 1609 // 2048*sqrt(2)*cos(5*pi/16)
	w6 = 1108 // 2048*sqrt(2)*cos(6*pi/16)
	w7 = 565  // 2048*sqrt(2)*cos(7*pi/16)

	w1pw7 = w1 + w7
	w1mw7 = w1 - w7
	w2pw6 = w2 + w6
	w2mw6 = w2 - w6
	w3pw5 = w3 + w5

	// standard deviation is sqrt( average( (x - xbar)^2 ) )
	// = sqrt( Integral( x^2 + xbar^2 -2*x*xbar dt ) / Duration )
	// = sqrt( ( Integral( x^2 dt ) + Duration * xbar^2 - 2*xbar*Integral(x dt) ) / Duration)
	// = sqrt( Integral(x^2 dt)/Duration - xbar^2 )
	variance := igr.IntegralXX/igr.ElapsedSeconds - avg*avg
	if variance >= 0 {
		return avg, math.Sqrt(variance)
	}
	return avg, math.NaN()

		// Safe to use J(n+1,x)=2n/x *J(n,x)-J(n-1,x)
		if x >= Two302 { // x > 2**302

			// (x >> n**2)
			//          Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
			//          Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
			//          Let s=sin(x), c=cos(x),
			//              xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
			//
			//                 n    sin(xn)*sqt2    cos(xn)*sqt2
			//              ----------------------------------

	MOVDW	F4, F5			// 85081b01
	NEGF	F4, F5			// 85141401
	NEGD	F4, F5			// 85181401
	ABSD	F4, F5			// 85081401
	TRUNCDW	F4, F5			// 85881a01
	TRUNCFW	F4, F5			// 85841a01
	SQRTF	F4, F5			// 85441401
	SQRTD	F4, F5			// 85481401

	DBAR	 			// 00007238
	NOOP	 			// 00004003

	MOVWR	R4, result+16(FP) 	// 6460402f
	MOVWR	R4, 1(R5) 		// a404402f
	MOVWR	y+8(FP), R4 		// 6440402e

}

// chunkSizeFor returns a chunk size for the given number of items to use for
// parallel work. The size aims to produce good CPU utilization.
// returns max(1, min(sqrt(n), n/Parallelism))
func chunkSizeFor(n, parallelism int) int {
	s := int(math.Sqrt(float64(n)))

	if r := n/parallelism + 1; s > r {
		s = r
	} else if s < 1 {
		s = 1
	}
	return s
}

	FDIVD   F1, F0       // F0=q(=q/p), F1=p
	FMULD   F0, F0       // F0=q*q, F1=p
	FLD1                 // F0=1, F1=q*q, F2=p
	FADDDP  F0, F1       // F0=1+q*q, F1=p
	FSQRT                // F0=sqrt(1+q*q), F1=p
	FMULDP  F0, F1       // F0=p*sqrt(1+q*q)
	FMOVDP  F0, ret+16(FP)
	RET
	FMOVDP  F0, F1       // F0=0
	FMOVDP  F0, ret+16(FP)
	RET
not_finite:
// test bits for -Inf or +Inf
	MOVL    p_hi+4(FP), AX  // high word p

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package p

import "math"

func f() {
	_ = min(0.1, 0.2, math.Sqrt(1))

		return complex(math.Copysign(math.Pi/2, re), math.Copysign(re, im))
	}
	ct := complex(-imag(x), real(x)) // i * x
	xx := x * x
	x1 := complex(1-real(xx), -imag(xx)) // 1 - x*x
	x2 := Sqrt(x1)                       // x2 = sqrt(1 - x*x)
	w := Log(ct + x2)
	return complex(imag(w), -real(w)) // -i * w
}

// Asinh returns the inverse hyperbolic sine of x.
func Asinh(x complex128) complex128 {

      int j = i & ARRAY_MASK;
      tmp += LongMath.log10(positive[j], mode);
    }
    return tmp;
  }

  @Benchmark
  int sqrt(int reps) {
    int tmp = 0;
    for (int i = 0; i < reps; i++) {
      int j = i & ARRAY_MASK;
      tmp += LongMath.sqrt(positive[j], mode);
    }
    return tmp;
  }

  @Benchmark
  int divide(int reps) {
    int tmp = 0;
    for (int i = 0; i < reps; i++) {

      int j = i & ARRAY_MASK;
      tmp += IntMath.log10(positive[j], mode);
    }
    return tmp;
  }

  @Benchmark
  int sqrt(int reps) {
    int tmp = 0;
    for (int i = 0; i < reps; i++) {
      int j = i & ARRAY_MASK;
      tmp += IntMath.sqrt(positive[j], mode);
    }
    return tmp;
  }

  @Benchmark
  int divide(int reps) {
    int tmp = 0;
    for (int i = 0; i < reps; i++) {