xref: /linux/arch/mips/math-emu/sp_mul.c (revision 93d90ad708b8da6efc0e487b66111aa9db7f70c7)
1 /* IEEE754 floating point arithmetic
2  * single precision
3  */
4 /*
5  * MIPS floating point support
6  * Copyright (C) 1994-2000 Algorithmics Ltd.
7  *
8  *  This program is free software; you can distribute it and/or modify it
9  *  under the terms of the GNU General Public License (Version 2) as
10  *  published by the Free Software Foundation.
11  *
12  *  This program is distributed in the hope it will be useful, but WITHOUT
13  *  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
14  *  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
15  *  for more details.
16  *
17  *  You should have received a copy of the GNU General Public License along
18  *  with this program; if not, write to the Free Software Foundation, Inc.,
19  *  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA.
20  */
21 
22 #include "ieee754sp.h"
23 
24 union ieee754sp ieee754sp_mul(union ieee754sp x, union ieee754sp y)
25 {
26 	int re;
27 	int rs;
28 	unsigned rm;
29 	unsigned short lxm;
30 	unsigned short hxm;
31 	unsigned short lym;
32 	unsigned short hym;
33 	unsigned lrm;
34 	unsigned hrm;
35 	unsigned t;
36 	unsigned at;
37 
38 	COMPXSP;
39 	COMPYSP;
40 
41 	EXPLODEXSP;
42 	EXPLODEYSP;
43 
44 	ieee754_clearcx();
45 
46 	FLUSHXSP;
47 	FLUSHYSP;
48 
49 	switch (CLPAIR(xc, yc)) {
50 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_QNAN):
51 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_SNAN):
52 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_SNAN):
53 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_SNAN):
54 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_SNAN):
55 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_SNAN):
56 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_SNAN):
57 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_ZERO):
58 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_NORM):
59 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_DNORM):
60 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_INF):
61 		ieee754_setcx(IEEE754_INVALID_OPERATION);
62 		return ieee754sp_nanxcpt(ieee754sp_indef());
63 
64 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_QNAN):
65 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_QNAN):
66 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_QNAN):
67 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_QNAN):
68 		return y;
69 
70 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_QNAN):
71 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_ZERO):
72 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_NORM):
73 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_DNORM):
74 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_INF):
75 		return x;
76 
77 
78 	/*
79 	 * Infinity handling
80 	 */
81 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_ZERO):
82 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_INF):
83 		ieee754_setcx(IEEE754_INVALID_OPERATION);
84 		return ieee754sp_indef();
85 
86 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_INF):
87 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_INF):
88 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM):
89 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM):
90 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF):
91 		return ieee754sp_inf(xs ^ ys);
92 
93 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO):
94 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM):
95 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_DNORM):
96 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_ZERO):
97 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_ZERO):
98 		return ieee754sp_zero(xs ^ ys);
99 
100 
101 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM):
102 		SPDNORMX;
103 
104 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM):
105 		SPDNORMY;
106 		break;
107 
108 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM):
109 		SPDNORMX;
110 		break;
111 
112 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM):
113 		break;
114 	}
115 	/* rm = xm * ym, re = xe+ye basically */
116 	assert(xm & SP_HIDDEN_BIT);
117 	assert(ym & SP_HIDDEN_BIT);
118 
119 	re = xe + ye;
120 	rs = xs ^ ys;
121 
122 	/* shunt to top of word */
123 	xm <<= 32 - (SP_FBITS + 1);
124 	ym <<= 32 - (SP_FBITS + 1);
125 
126 	/*
127 	 * Multiply 32 bits xm, ym to give high 32 bits rm with stickness.
128 	 */
129 	lxm = xm & 0xffff;
130 	hxm = xm >> 16;
131 	lym = ym & 0xffff;
132 	hym = ym >> 16;
133 
134 	lrm = lxm * lym;	/* 16 * 16 => 32 */
135 	hrm = hxm * hym;	/* 16 * 16 => 32 */
136 
137 	t = lxm * hym; /* 16 * 16 => 32 */
138 	at = lrm + (t << 16);
139 	hrm += at < lrm;
140 	lrm = at;
141 	hrm = hrm + (t >> 16);
142 
143 	t = hxm * lym; /* 16 * 16 => 32 */
144 	at = lrm + (t << 16);
145 	hrm += at < lrm;
146 	lrm = at;
147 	hrm = hrm + (t >> 16);
148 
149 	rm = hrm | (lrm != 0);
150 
151 	/*
152 	 * Sticky shift down to normal rounding precision.
153 	 */
154 	if ((int) rm < 0) {
155 		rm = (rm >> (32 - (SP_FBITS + 1 + 3))) |
156 		    ((rm << (SP_FBITS + 1 + 3)) != 0);
157 		re++;
158 	} else {
159 		rm = (rm >> (32 - (SP_FBITS + 1 + 3 + 1))) |
160 		     ((rm << (SP_FBITS + 1 + 3 + 1)) != 0);
161 	}
162 	assert(rm & (SP_HIDDEN_BIT << 3));
163 
164 	return ieee754sp_format(rs, re, rm);
165 }
166