Annotation of 43BSDReno/lib/libm/mc68881/atan2.c, revision 1.1.1.1

1.1       root        1: /*-
                      2:  * Copyright (c) 1990 The Regents of the University of California.
                      3:  * All rights reserved.
                      4:  *
                      5:  * This code is derived from software contributed to Berkeley by
                      6:  * the Systems Programming Group of the University of Utah Computer
                      7:  * Science Department.
                      8:  *
                      9:  * Redistribution and use in source and binary forms are permitted
                     10:  * provided that: (1) source distributions retain this entire copyright
                     11:  * notice and comment, and (2) distributions including binaries display
                     12:  * the following acknowledgement:  ``This product includes software
                     13:  * developed by the University of California, Berkeley and its contributors''
                     14:  * in the documentation or other materials provided with the distribution
                     15:  * and in all advertising materials mentioning features or use of this
                     16:  * software. Neither the name of the University nor the names of its
                     17:  * contributors may be used to endorse or promote products derived
                     18:  * from this software without specific prior written permission.
                     19:  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
                     20:  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
                     21:  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
                     22:  */
                     23: 
                     24: #ifndef lint
                     25: static char sccsid[] = "@(#)atan2.c    5.1 (Berkeley) 5/17/90";
                     26: #endif /* not lint */
                     27: 
                     28: /*
                     29:  * ATAN2(Y,X)
                     30:  * RETURN ARG (X+iY)
                     31:  * DOUBLE PRECISION (IEEE DOUBLE 53 BITS)
                     32:  *
                     33:  * Scaled down version to weed out special cases.  "Normal" cases are
                     34:  * handled by calling atan2__A(), an assembly coded support routine in
                     35:  * support.s.
                     36:  *
                     37:  * Required system supported functions :
                     38:  *     copysign(x,y)
                     39:  *     atan2__A(y,x)
                     40:  *     
                     41:  * Method :
                     42:  *     1. Deal with special cases
                     43:  *     2. Call atan2__A() to do the others
                     44:  *
                     45:  * Special cases:
                     46:  * Notations: atan2(y,x) == ARG (x+iy) == ARG(x,y).
                     47:  *
                     48:  *     ARG( NAN , (anything) ) is NaN;
                     49:  *     ARG( (anything), NaN ) is NaN;
                     50:  *     ARG(+(anything but NaN), +-0) is +-0  ;
                     51:  *     ARG(-(anything but NaN), +-0) is +-PI ;
                     52:  *     ARG( 0, +-(anything but 0 and NaN) ) is +-PI/2;
                     53:  *     ARG( +INF,+-(anything but INF and NaN) ) is +-0 ;
                     54:  *     ARG( -INF,+-(anything but INF and NaN) ) is +-PI;
                     55:  *     ARG( +INF,+-INF ) is +-PI/4 ;
                     56:  *     ARG( -INF,+-INF ) is +-3PI/4;
                     57:  *     ARG( (anything but,0,NaN, and INF),+-INF ) is +-PI/2;
                     58:  *
                     59:  * Accuracy:
                     60:  *     atan2(y,x) returns (PI/pi) * the exact ARG (x+iy) nearly rounded, 
                     61:  *     where
                     62:  *
                     63:  *     in decimal:
                     64:  *             pi = 3.141592653589793 23846264338327 ..... 
                     65:  *    53 bits   PI = 3.141592653589793 115997963 ..... ,
                     66:  *    56 bits   PI = 3.141592653589793 227020265 ..... ,  
                     67:  *
                     68:  *     in hexadecimal:
                     69:  *             pi = 3.243F6A8885A308D313198A2E....
                     70:  *    53 bits   PI = 3.243F6A8885A30  =  2 * 1.921FB54442D18   error=.276ulps
                     71:  *    56 bits   PI = 3.243F6A8885A308 =  4 * .C90FDAA22168C2    error=.206ulps
                     72:  *     
                     73:  *     In a test run with 356,000 random argument on [-1,1] * [-1,1] on a
                     74:  *     VAX, the maximum observed error was 1.41 ulps (units of the last place)
                     75:  *     compared with (PI/pi)*(the exact ARG(x+iy)).
                     76:  *
                     77:  * Note:
                     78:  *     We use machine PI (the true pi rounded) in place of the actual
                     79:  *     value of pi for all the trig and inverse trig functions. In general, 
                     80:  *     if trig is one of sin, cos, tan, then computed trig(y) returns the 
                     81:  *     exact trig(y*pi/PI) nearly rounded; correspondingly, computed arctrig 
                     82:  *     returns the exact arctrig(y)*PI/pi nearly rounded. These guarantee the 
                     83:  *     trig functions have period PI, and trig(arctrig(x)) returns x for
                     84:  *     all critical values x.
                     85:  *     
                     86:  * Constants:
                     87:  * The hexadecimal values are the intended ones for the following constants.
                     88:  * The decimal values may be used, provided that the compiler will convert
                     89:  * from decimal to binary accurately enough to produce the hexadecimal values
                     90:  * shown.
                     91:  */
                     92: 
                     93: static double 
                     94: PIo4   =  7.8539816339744827900E-1    , /*Hex  2^ -1   *  1.921FB54442D18 */
                     95: PIo2   =  1.5707963267948965580E0     , /*Hex  2^  0   *  1.921FB54442D18 */
                     96: PI     =  3.1415926535897931160E0     ; /*Hex  2^  1   *  1.921FB54442D18 */
                     97: 
                     98: double atan2(y,x)
                     99: double  y,x;
                    100: {  
                    101:        static double zero=0, one=1;
                    102:        double copysign(),atan2__A(),signy,signx;
                    103:        int finite();
                    104: 
                    105:     /* if x or y is NAN */
                    106:        if(x!=x) return(x); if(y!=y) return(y);
                    107: 
                    108:     /* copy down the sign of y and x */
                    109:        signy = copysign(one,y);
                    110:        signx = copysign(one,x);
                    111: 
                    112:     /* when y = 0 */
                    113:        if(y==zero) return((signx==one)?y:copysign(PI,signy));
                    114: 
                    115:     /* when x = 0 */
                    116:        if(x==zero) return(copysign(PIo2,signy));
                    117:            
                    118:     /* when x is INF */
                    119:        if(!finite(x))
                    120:            if(!finite(y)) 
                    121:                return(copysign((signx==one)?PIo4:3*PIo4,signy));
                    122:            else
                    123:                return(copysign((signx==one)?zero:PI,signy));
                    124: 
                    125:     /* when y is INF */
                    126:        if(!finite(y)) return(copysign(PIo2,signy));
                    127: 
                    128:     /* else let atan2__A do the work */
                    129:        return(atan2__A(y,x));
                    130: }

unix.superglobalmegacorp.com

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.