xref: /freebsd/lib/msun/src/e_fmodf.c (revision 62a52c15422470f97fc7b311d89c83f910bcc1b1)
1 /* e_fmodf.c -- float version of e_fmod.c.
2  * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
3  */
4 
5 /*
6  * ====================================================
7  * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
8  *
9  * Developed at SunPro, a Sun Microsystems, Inc. business.
10  * Permission to use, copy, modify, and distribute this
11  * software is freely granted, provided that this notice
12  * is preserved.
13  * ====================================================
14  */
15 
16 /*
17  * fmodf(x,y)
18  * Return x mod y in exact arithmetic
19  * Method: shift and subtract
20  */
21 
22 #include "math.h"
23 #include "math_private.h"
24 
25 static const float one = 1.0, Zero[] = {0.0, -0.0,};
26 
27 float
28 fmodf(float x, float y)
29 {
30 	int32_t n,hx,hy,hz,ix,iy,sx,i;
31 
32 	GET_FLOAT_WORD(hx,x);
33 	GET_FLOAT_WORD(hy,y);
34 	sx = hx&0x80000000;		/* sign of x */
35 	hx ^=sx;		/* |x| */
36 	hy &= 0x7fffffff;	/* |y| */
37 
38     /* purge off exception values */
39 	if(hy==0||(hx>=0x7f800000)||		/* y=0,or x not finite */
40 	   (hy>0x7f800000))			/* or y is NaN */
41 	    return nan_mix_op(x, y, *)/nan_mix_op(x, y, *);
42 	if(hx<hy) return x;			/* |x|<|y| return x */
43 	if(hx==hy)
44 	    return Zero[(u_int32_t)sx>>31];	/* |x|=|y| return x*0*/
45 
46     /* determine ix = ilogb(x) */
47 	if(hx<0x00800000) {	/* subnormal x */
48 	    for (ix = -126,i=(hx<<8); i>0; i<<=1) ix -=1;
49 	} else ix = (hx>>23)-127;
50 
51     /* determine iy = ilogb(y) */
52 	if(hy<0x00800000) {	/* subnormal y */
53 	    for (iy = -126,i=(hy<<8); i>=0; i<<=1) iy -=1;
54 	} else iy = (hy>>23)-127;
55 
56     /* set up {hx,lx}, {hy,ly} and align y to x */
57 	if(ix >= -126)
58 	    hx = 0x00800000|(0x007fffff&hx);
59 	else {		/* subnormal x, shift x to normal */
60 	    n = -126-ix;
61 	    hx = hx<<n;
62 	}
63 	if(iy >= -126)
64 	    hy = 0x00800000|(0x007fffff&hy);
65 	else {		/* subnormal y, shift y to normal */
66 	    n = -126-iy;
67 	    hy = hy<<n;
68 	}
69 
70     /* fix point fmod */
71 	n = ix - iy;
72 	while(n--) {
73 	    hz=hx-hy;
74 	    if(hz<0){hx = hx+hx;}
75 	    else {
76 	    	if(hz==0) 		/* return sign(x)*0 */
77 		    return Zero[(u_int32_t)sx>>31];
78 	    	hx = hz+hz;
79 	    }
80 	}
81 	hz=hx-hy;
82 	if(hz>=0) {hx=hz;}
83 
84     /* convert back to floating value and restore the sign */
85 	if(hx==0) 			/* return sign(x)*0 */
86 	    return Zero[(u_int32_t)sx>>31];
87 	while(hx<0x00800000) {		/* normalize x */
88 	    hx = hx+hx;
89 	    iy -= 1;
90 	}
91 	if(iy>= -126) {		/* normalize output */
92 	    hx = ((hx-0x00800000)|((iy+127)<<23));
93 	    SET_FLOAT_WORD(x,hx|sx);
94 	} else {		/* subnormal output */
95 	    n = -126 - iy;
96 	    hx >>= n;
97 	    SET_FLOAT_WORD(x,hx|sx);
98 	    x *= one;		/* create necessary signal */
99 	}
100 	return x;		/* exact output */
101 }
102