xref: /linux/arch/mips/math-emu/sp_mul.c (revision bd628c1bed7902ec1f24ba0fe70758949146abbe)
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 		/* fall through */
105 
106 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM):
107 		SPDNORMY;
108 		break;
109 
110 	case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM):
111 		SPDNORMX;
112 		break;
113 
114 	case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM):
115 		break;
116 	}
117 	/* rm = xm * ym, re = xe+ye basically */
118 	assert(xm & SP_HIDDEN_BIT);
119 	assert(ym & SP_HIDDEN_BIT);
120 
121 	re = xe + ye;
122 	rs = xs ^ ys;
123 
124 	/* shunt to top of word */
125 	xm <<= 32 - (SP_FBITS + 1);
126 	ym <<= 32 - (SP_FBITS + 1);
127 
128 	/*
129 	 * Multiply 32 bits xm, ym to give high 32 bits rm with stickness.
130 	 */
131 	lxm = xm & 0xffff;
132 	hxm = xm >> 16;
133 	lym = ym & 0xffff;
134 	hym = ym >> 16;
135 
136 	lrm = lxm * lym;	/* 16 * 16 => 32 */
137 	hrm = hxm * hym;	/* 16 * 16 => 32 */
138 
139 	t = lxm * hym; /* 16 * 16 => 32 */
140 	at = lrm + (t << 16);
141 	hrm += at < lrm;
142 	lrm = at;
143 	hrm = hrm + (t >> 16);
144 
145 	t = hxm * lym; /* 16 * 16 => 32 */
146 	at = lrm + (t << 16);
147 	hrm += at < lrm;
148 	lrm = at;
149 	hrm = hrm + (t >> 16);
150 
151 	rm = hrm | (lrm != 0);
152 
153 	/*
154 	 * Sticky shift down to normal rounding precision.
155 	 */
156 	if ((int) rm < 0) {
157 		rm = (rm >> (32 - (SP_FBITS + 1 + 3))) |
158 		    ((rm << (SP_FBITS + 1 + 3)) != 0);
159 		re++;
160 	} else {
161 		rm = (rm >> (32 - (SP_FBITS + 1 + 3 + 1))) |
162 		     ((rm << (SP_FBITS + 1 + 3 + 1)) != 0);
163 	}
164 	assert(rm & (SP_HIDDEN_BIT << 3));
165 
166 	return ieee754sp_format(rs, re, rm);
167 }
168