/*	$OpenBSD: n_lgamma.c,v 1.9 2009/10/27 23:59:29 deraadt Exp $	*/
/*-
 * Copyright (c) 1992, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

/*
 * Coded by Peter McIlroy, Nov 1992;
 *
 * The financial support of UUNET Communications Services is greatfully
 * acknowledged.
 */

#include <math.h>
#include <errno.h>

#include "mathimpl.h"

/* Log gamma function.
 * Error:  x > 0 error < 1.3ulp.
 *	   x > 4, error < 1ulp.
 *	   x > 9, error < .6ulp.
 * 	   x < 0, all bets are off. (When G(x) ~ 1, log(G(x)) ~ 0)
 * Method:
 *	x > 6:
 *		Use the asymptotic expansion (Stirling's Formula)
 *	0 < x < 6:
 *		Use tgamma(x+1) = x*tgamma(x) for argument reduction.
 *		Use rational approximation in
 *		the range 1.2, 2.5
 *		Two approximations are used, one centered at the
 *		minimum to ensure monotonicity; one centered at 2
 *		to maintain small relative error.
 *	x < 0:
 *		Use the reflection formula,
 *		G(1-x)G(x) = PI/sin(PI*x)
 * Special values:
 *	non-positive integer	returns +Inf.
 *	NaN			returns NaN
*/
#if defined(__vax__)
#define _IEEE		0
/* double and float have same size exponent field */
#define TRUNC(x)	x = (double) (float) (x)
#else
static int endian;
#define _IEEE		1
#define TRUNC(x)	*(((int *) &x) + endian) &= 0xf8000000
#define infnan(x)	0.0
#endif

static double small_lgam(double);
static double large_lgam(double);
static double neg_lgam(double);
static const double one = 1.0;
extern int signgam;

#define UNDERFL (1e-1020 * 1e-1020)

#define LEFT	(1.0 - (x0 + .25))
#define RIGHT	(x0 - .218)
/*
 * Constants for approximation in [1.244,1.712]
 */
#define x0	0.461632144968362356785
#define x0_lo	-.000000000000000015522348162858676890521
#define a0_hi	-0.12148629128932952880859
#define a0_lo	.0000000007534799204229502
#define r0	-2.771227512955130520e-002
#define r1	-2.980729795228150847e-001
#define r2	-3.257411333183093394e-001
#define r3	-1.126814387531706041e-001
#define r4	-1.129130057170225562e-002
#define r5	-2.259650588213369095e-005
#define s0	 1.714457160001714442e+000
#define s1	 2.786469504618194648e+000
#define s2	 1.564546365519179805e+000
#define s3	 3.485846389981109850e-001
#define s4	 2.467759345363656348e-002
/*
 * Constants for approximation in [1.71, 2.5]
*/
#define a1_hi	4.227843350984671344505727574870e-01
#define a1_lo	4.670126436531227189e-18
#define p0	3.224670334241133695662995251041e-01
#define p1	3.569659696950364669021382724168e-01
#define p2	1.342918716072560025853732668111e-01
#define p3	1.950702176409779831089963408886e-02
#define p4	8.546740251667538090796227834289e-04
#define q0	1.000000000000000444089209850062e+00
#define q1	1.315850076960161985084596381057e+00
#define q2	6.274644311862156431658377186977e-01
#define q3	1.304706631926259297049597307705e-01
#define q4	1.102815279606722369265536798366e-02
#define q5	2.512690594856678929537585620579e-04
#define q6	-1.003597548112371003358107325598e-06
/*
 * Stirling's Formula, adjusted for equal-ripple. x in [6,Inf].
*/
#define lns2pi	.418938533204672741780329736405
#define pb0	 8.33333333333333148296162562474e-02
#define pb1	-2.77777777774548123579378966497e-03
#define pb2	 7.93650778754435631476282786423e-04
#define pb3	-5.95235082566672847950717262222e-04
#define pb4	 8.41428560346653702135821806252e-04
#define pb5	-1.89773526463879200348872089421e-03
#define pb6	 5.69394463439411649408050664078e-03
#define pb7	-1.44705562421428915453880392761e-02

__pure double
lgamma(double x)
{
	double r;

	int signgam = 1;
#if _IEEE
	endian = ((*(int *) &one)) ? 1 : 0;
#endif

	if (!finite(x))
		if (_IEEE)
			return (x+x);
		else return (infnan(EDOM));

	if (x > 6 + RIGHT) {
		r = large_lgam(x);
		return (r);
	} else if (x > 1e-16)
		return (small_lgam(x));
	else if (x > -1e-16) {
		if (x < 0) {
			signgam = -1;
			x = -x;
		}
		return (-log(x));
	} else
		return (neg_lgam(x));
}

float
lgammaf(float x)
{
	return lgamma(x);
}

/*
 * The gamma() function performs identically to lgamma(), including
 * the use of signgam.
 */

double
gamma(double x)
{
	return lgamma(x);
}

float
gammaf(float x)
{
	return lgammaf(x);
}

static double
large_lgam(double x)
{
	double z, p, x1;
	struct Double t, u, v;
	u = __log__D(x);
	u.a -= 1.0;
	if (x > 1e15) {
		v.a = x - 0.5;
		TRUNC(v.a);
		v.b = (x - v.a) - 0.5;
		t.a = u.a*v.a;
		t.b = x*u.b + v.b*u.a;
		if (_IEEE == 0 && !finite(t.a))
			return(infnan(ERANGE));
		return(t.a + t.b);
	}
	x1 = 1./x;
	z = x1*x1;
	p = pb0+z*(pb1+z*(pb2+z*(pb3+z*(pb4+z*(pb5+z*(pb6+z*pb7))))));
					/* error in approximation = 2.8e-19 */

	p = p*x1;			/* error < 2.3e-18 absolute */
					/* 0 < p < 1/64 (at x = 5.5) */
	v.a = x = x - 0.5;
	TRUNC(v.a);			/* truncate v.a to 26 bits. */
	v.b = x - v.a;
	t.a = v.a*u.a;			/* t = (x-.5)*(log(x)-1) */
	t.b = v.b*u.a + x*u.b;
	t.b += p; t.b += lns2pi;	/* return t + lns2pi + p */
	return (t.a + t.b);
}

static double
small_lgam(double x)
{
	int x_int;
	double y, z, t, r = 0, p, q, hi, lo;
	struct Double rr;
	x_int = (x + .5);
	y = x - x_int;
	if (x_int <= 2 && y > RIGHT) {
		t = y - x0;
		y--; x_int++;
		goto CONTINUE;
	} else if (y < -LEFT) {
		t = y +(1.0-x0);
CONTINUE:
		z = t - x0_lo;
		p = r0+z*(r1+z*(r2+z*(r3+z*(r4+z*r5))));
		q = s0+z*(s1+z*(s2+z*(s3+z*s4)));
		r = t*(z*(p/q) - x0_lo);
		t = .5*t*t;
		z = 1.0;
		switch (x_int) {
		case 6:	z  = (y + 5);		/* FALLTHROUGH */
		case 5:	z *= (y + 4);		/* FALLTHROUGH */
		case 4:	z *= (y + 3);		/* FALLTHROUGH */
		case 3:	z *= (y + 2);
			rr = __log__D(z);
			rr.b += a0_lo; rr.a += a0_hi;
			return(((r+rr.b)+t+rr.a));
		case 2: return(((r+a0_lo)+t)+a0_hi);
		case 0: r -= log1p(x);	/* FALLTHROUGH */
		default: rr = __log__D(x);
			rr.a -= a0_hi; rr.b -= a0_lo;
			return(((r - rr.b) + t) - rr.a);
		}
	} else {
		p = p0+y*(p1+y*(p2+y*(p3+y*p4)));
		q = q0+y*(q1+y*(q2+y*(q3+y*(q4+y*(q5+y*q6)))));
		p = p*(y/q);
		t = (double)(float) y;
		z = y-t;
		hi = (double)(float) (p+a1_hi);
		lo = a1_hi - hi; lo += p; lo += a1_lo;
		r = lo*y + z*hi;	/* q + r = y*(a0+p/q) */
		q = hi*t;
		z = 1.0;
		switch (x_int) {
		case 6:	z  = (y + 5);		/* FALLTHROUGH */
		case 5:	z *= (y + 4);		/* FALLTHROUGH */
		case 4:	z *= (y + 3);		/* FALLTHROUGH */
		case 3:	z *= (y + 2);		/* FALLTHROUGH */
			rr = __log__D(z);
			r += rr.b; r += q;
			return(rr.a + r);
		case 2:	return (q+ r);
		case 0: rr = __log__D(x);
			r -= rr.b; r -= log1p(x);
			r += q; r-= rr.a;
			return(r);
		default: rr = __log__D(x);
			r -= rr.b;
			q -= rr.a;
			return (r+q);
		}
	}
}

static double
neg_lgam(double x)
{
	int xi;
	double y, z, zero = 0.0;

	/* avoid destructive cancellation as much as possible */
	if (x > -170) {
		xi = x;
		if (xi == x)
			if (_IEEE)
				return(one/zero);
			else
				return(infnan(ERANGE));
		y = tgamma(x);
		if (y < 0) {
			y = -y;
			signgam = -1;
		}
		return (log(y));
	}
	z = floor(x + .5);
	if (z == x) {		/* convention: G(-(integer)) -> +Inf */
		if (_IEEE)
			return (one/zero);
		else
			return (infnan(ERANGE));
	}
	y = .5*ceil(x);
	if (y == ceil(y))
		signgam = -1;
	x = -x;
	z = fabs(x + z);	/* 0 < z <= .5 */
	if (z < .25)
		z = sin(M_PI*z);
	else
		z = cos(M_PI*(0.5-z));
	z = log(M_PI/(z*x));
	y = large_lgam(x);
	return (z - y);
}