s_expm1.c revision 6b15695578f07a3f72c4c9475c1a261a3021472a
70N/A/* -*- Mode: C; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*- 1276N/A * ***** BEGIN LICENSE BLOCK ***** 70N/A * Version: MPL 1.1/GPL 2.0/LGPL 2.1 70N/A * The contents of this file are subject to the Mozilla Public License Version 70N/A * 1.1 (the "License"); you may not use this file except in compliance with 919N/A * the License. You may obtain a copy of the License at 919N/A * Software distributed under the License is distributed on an "AS IS" basis, 919N/A * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License 919N/A * for the specific language governing rights and limitations under the 919N/A * The Original Code is Mozilla Communicator client code, released 919N/A * The Initial Developer of the Original Code is 919N/A * Sun Microsystems, Inc. 919N/A * Portions created by the Initial Developer are Copyright (C) 1998 919N/A * the Initial Developer. All Rights Reserved. 70N/A * Alternatively, the contents of this file may be used under the terms of 70N/A * either of the GNU General Public License Version 2 or later (the "GPL"), 70N/A * or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"), 493N/A * in which case the provisions of the GPL or the LGPL are applicable instead 70N/A * of those above. If you wish to allow use of your version of this file only 70N/A * under the terms of either the GPL or the LGPL, and not to allow others to 1408N/A * use your version of this file under the terms of the MPL, indicate your 70N/A * decision by deleting the provisions above and replace them with the notice 911N/A * and other provisions required by the GPL or the LGPL. If you do not delete 1408N/A * the provisions above, a recipient may use your version of this file under 1408N/A * the terms of any one of the MPL, the GPL or the LGPL. 911N/A * ***** END LICENSE BLOCK ***** */ 70N/A * ==================================================== 70N/A * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. 970N/A * Developed at SunSoft, a Sun Microsystems, Inc. business. 970N/A * Permission to use, copy, modify, and distribute this 970N/A * software is freely granted, provided that this notice 70N/A * ==================================================== 70N/A * Returns exp(x)-1, the exponential of x minus 1. 70N/A * 1. Argument reduction: 970N/A * Given x, find r and integer k such that 970N/A * x = k*ln2 + r, |r| <= 0.5*ln2 ~ 0.34658 * Here a correction term c will be computed to compensate * the error in r when rounded to a floating-point number. * 2. Approximating expm1(r) by a special rational function on * the interval [0,0.34658]: * r*(exp(r)+1)/(exp(r)-1) = 2+ r^2/6 - r^4/360 + ... * r*(exp(r)+1)/(exp(r)-1) = 2+ r^2/6 * R1(r*r) * R1(r**2) = 6/r *((exp(r)+1)/(exp(r)-1) - 2/r) * = 6/r * ( 1 + 2.0*(1/(exp(r)-1) - 1/r)) * = 1 - r^2/60 + r^4/2520 - r^6/100800 + ... * We use a special Reme algorithm on [0,0.347] to generate * a polynomial of degree 5 in r*r to approximate R1. The * maximum error of this polynomial approximation is bounded * by 2**-61. In other words, * R1(z) ~ 1.0 + Q1*z + Q2*z**2 + Q3*z**3 + Q4*z**4 + Q5*z**5 * where Q1 = -1.6666666666666567384E-2, * Q2 = 3.9682539681370365873E-4, * Q3 = -9.9206344733435987357E-6, * Q4 = 2.5051361420808517002E-7, * Q5 = -6.2843505682382617102E-9; * (where z=r*r, and the values of Q1 to Q5 are listed below) * | 1.0+Q1*z+...+Q5*z - R1(z) | <= 2 * expm1(r) = exp(r)-1 is then computed by the following * specific way which minimize the accumulation rounding error: * r r [ 3 - (R1 + R1*r/2) ] * expm1(r) = r + --- + --- * [--------------------] * 2 2 [ 6 - r*(3 - R1*r/2) ] * To compensate the error in the argument reduction, we use * expm1(r+c) = expm1(r) + c + expm1(r)*c * Thus c+r*c will be added in as the correction terms for * expm1(r+c). Now rearrange the term to avoid optimization * ({ ( r [ R1 - (3 - R1*r/2) ] ) } r ) * expm1(r+c)~r - ({r*(--- * [--------------------]-c)-c} - --- ) * ({ ( 2 [ 6 - r*(3 - R1*r/2) ] ) } 2 ) * 3. Scale back to obtain expm1(x): * expm1(x) = either 2^k*[expm1(r)+1] - 1 * = or 2^k*[expm1(r) + (1-2^-k)] * 4. Implementation notes: * (A). To save one multiplication, we scale the coefficient Qi * to Qi*2^i, and replace z by (x^2)/2. * (B). To achieve maximum accuracy, we compute expm1(x) by * (i) if x < -56*ln2, return -1.0, (raise inexact if x!=inf) * (ii) if k=0, return r-E * (iii) if k=-1, return 0.5*(r-E)-0.5 * (iv) if k=1 if r < -0.25, return 2*((r+0.5)- E) * else return 1.0+2.0*(r-E); * (v) if (k<-2||k>56) return 2^k(1-(E-r)) - 1 (or exp(x)-1) * (vi) if k <= 20, return 2^k((1-2^-k)-(E-r)), else * (vii) return 2^k(1-((E+2^-k)-r)) * expm1(INF) is INF, expm1(NaN) is NaN; * for finite argument, only expm1(0)=0 is exact. * according to an error analysis, the error is always less than * 1 ulp (unit in the last place). * if x > 7.09782712893383973096e+02 then expm1(x) overflow * The hexadecimal values are the intended ones for the following * constants. The decimal values may be used, provided that the * compiler will convert from decimal to binary accurately enough * to produce the hexadecimal values shown. o_threshold =
7.09782712893383973096e+02,
/* 0x40862E42, 0xFEFA39EF */ ln2_hi =
6.93147180369123816490e-01,
/* 0x3fe62e42, 0xfee00000 */ ln2_lo =
1.90821492927058770002e-10,
/* 0x3dea39ef, 0x35793c76 */ invln2 =
1.44269504088896338700e+00,
/* 0x3ff71547, 0x652b82fe */ /* scaled coefficients related to expm1 */ Q1 = -
3.33333333333331316428e-02,
/* BFA11111 111110F4 */ Q2 =
1.58730158725481460165e-03,
/* 3F5A01A0 19FE5585 */ Q3 = -
7.93650757867487942473e-05,
/* BF14CE19 9EAADBB7 */ Q4 =
4.00821782732936239552e-06,
/* 3ED0CFCA 86E65239 */ Q5 = -
2.01099218183624371326e-07;
/* BE8AFDB7 6E09C32D */ hx =
__HI(u);
/* high word of x */ xsb =
hx&
0x80000000;
/* sign bit of x */ if(
xsb==0) y=x;
else y= -x;
/* y = |x| */ hx &=
0x7fffffff;
/* high word of |x| */ /* filter out huge and non-finite argument */ if(
hx >=
0x4043687A) {
/* if |x|>=56*ln2 */ if(
hx >=
0x40862E42) {
/* if |x|>=709.78... */ else return (
xsb==0)? x:-
1.0;
/* exp(+-inf)={inf,-1} */ if(
xsb!=0) {
/* x < -56*ln2, return -1.0 with inexact */ if(x+
tiny<
0.0)
/* raise inexact */ if(
hx >
0x3fd62e42) {
/* if |x| > 0.5 ln2 */ if(
hx <
0x3FF0A2B2) {
/* and |x| < 1.5 ln2 */ hi = x - t*
ln2_hi;
/* t*ln2_hi is exact here */ else if(
hx <
0x3c900000) {
/* when |x|<2**-54, return x */ t =
really_big+x;
/* return x with inexact flags when x!=0 */ /* x is now in primary range */ e =
hxs*((
r1-t)/(
6.0 - x*t));
if(k==0)
return x - (x*e-
hxs);
/* c is 0 */ if(k== -
1)
return 0.5*(x-e)-
0.5;
if(x < -
0.25)
return -
2.0*(e-(x+
0.5));
else return one+
2.0*(x-e);
if (k <= -
2 || k>
56) {
/* suffice to return exp(x)-1 */ __HI(u) += (k<<
20);
/* add k to y's exponent */ __HI(u) =
0x3ff00000 - (
0x200000>>k);
/* t=1-2^-k */ __HI(u) += (k<<
20);
/* add k to y's exponent */ __HI(u) = ((
0x3ff-k)<<
20);
/* 2^-k */ __HI(u) += (k<<
20);
/* add k to y's exponent */