xref: /linux/arch/mips/math-emu/sp_mul.c (revision ef69f8d2ff09518657c3ecaf2db8408c16549829)
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 int rm;
29 	unsigned short lxm;
30 	unsigned short hxm;
31 	unsigned short lym;
32 	unsigned short hym;
33 	unsigned int lrm;
34 	unsigned int hrm;
35 	unsigned int t;
36 	unsigned int 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_QNAN, IEEE754_CLASS_SNAN):
51 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_SNAN):
52 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_SNAN):
53 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_SNAN):
54 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_SNAN):
55 		return ieee754sp_nanxcpt(y);
56 
57 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_SNAN):
58 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_QNAN):
59 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_ZERO):
60 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_NORM):
61 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_DNORM):
62 	case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_INF):
63 		return ieee754sp_nanxcpt(x);
64 
65 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_QNAN):
66 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_QNAN):
67 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_QNAN):
68 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_QNAN):
69 		return y;
70 
71 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_QNAN):
72 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_ZERO):
73 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_NORM):
74 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_DNORM):
75 	case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_INF):
76 		return x;
77 
78 
79 	/*
80 	 * Infinity handling
81 	 */
82 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_ZERO):
83 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_INF):
84 		ieee754_setcx(IEEE754_INVALID_OPERATION);
85 		return ieee754sp_indef();
86 
87 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_INF):
88 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_INF):
89 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM):
90 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM):
91 	case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF):
92 		return ieee754sp_inf(xs ^ ys);
93 
94 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO):
95 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM):
96 	case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_DNORM):
97 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_ZERO):
98 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_ZERO):
99 		return ieee754sp_zero(xs ^ ys);
100 
101 
102 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM):
103 		SPDNORMX;
104 
105 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM):
106 		SPDNORMY;
107 		break;
108 
109 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM):
110 		SPDNORMX;
111 		break;
112 
113 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM):
114 		break;
115 	}
116 	/* rm = xm * ym, re = xe+ye basically */
117 	assert(xm & SP_HIDDEN_BIT);
118 	assert(ym & SP_HIDDEN_BIT);
119 
120 	re = xe + ye;
121 	rs = xs ^ ys;
122 
123 	/* shunt to top of word */
124 	xm <<= 32 - (SP_FBITS + 1);
125 	ym <<= 32 - (SP_FBITS + 1);
126 
127 	/*
128 	 * Multiply 32 bits xm, ym to give high 32 bits rm with stickness.
129 	 */
130 	lxm = xm & 0xffff;
131 	hxm = xm >> 16;
132 	lym = ym & 0xffff;
133 	hym = ym >> 16;
134 
135 	lrm = lxm * lym;	/* 16 * 16 => 32 */
136 	hrm = hxm * hym;	/* 16 * 16 => 32 */
137 
138 	t = lxm * hym; /* 16 * 16 => 32 */
139 	at = lrm + (t << 16);
140 	hrm += at < lrm;
141 	lrm = at;
142 	hrm = hrm + (t >> 16);
143 
144 	t = hxm * lym; /* 16 * 16 => 32 */
145 	at = lrm + (t << 16);
146 	hrm += at < lrm;
147 	lrm = at;
148 	hrm = hrm + (t >> 16);
149 
150 	rm = hrm | (lrm != 0);
151 
152 	/*
153 	 * Sticky shift down to normal rounding precision.
154 	 */
155 	if ((int) rm < 0) {
156 		rm = (rm >> (32 - (SP_FBITS + 1 + 3))) |
157 		    ((rm << (SP_FBITS + 1 + 3)) != 0);
158 		re++;
159 	} else {
160 		rm = (rm >> (32 - (SP_FBITS + 1 + 3 + 1))) |
161 		     ((rm << (SP_FBITS + 1 + 3 + 1)) != 0);
162 	}
163 	assert(rm & (SP_HIDDEN_BIT << 3));
164 
165 	return ieee754sp_format(rs, re, rm);
166 }
167