tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* IEEE754 floating point arithmetic
2 * double precision: common utilities
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 "ieee754dp.h"
23
24union ieee754dp ieee754dp_mul(union ieee754dp x, union ieee754dp y)
25{
26 int re;
27 int rs;
28 u64 rm;
29 unsigned lxm;
30 unsigned hxm;
31 unsigned lym;
32 unsigned hym;
33 u64 lrm;
34 u64 hrm;
35 u64 t;
36 u64 at;
37
38 COMPXDP;
39 COMPYDP;
40
41 EXPLODEXDP;
42 EXPLODEYDP;
43
44 ieee754_clearcx();
45
46 FLUSHXDP;
47 FLUSHYDP;
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 ieee754dp_nanxcpt(ieee754dp_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 ieee754dp_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 ieee754dp_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 ieee754dp_zero(xs ^ ys);
99
100
101 case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM):
102 DPDNORMX;
103
104 case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM):
105 DPDNORMY;
106 break;
107
108 case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM):
109 DPDNORMX;
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 & DP_HIDDEN_BIT);
117 assert(ym & DP_HIDDEN_BIT);
118
119 re = xe + ye;
120 rs = xs ^ ys;
121
122 /* shunt to top of word */
123 xm <<= 64 - (DP_FBITS + 1);
124 ym <<= 64 - (DP_FBITS + 1);
125
126 /*
127 * Multiply 32 bits xm, ym to give high 32 bits rm with stickness.
128 */
129
130 /* 32 * 32 => 64 */
131#define DPXMULT(x, y) ((u64)(x) * (u64)y)
132
133 lxm = xm;
134 hxm = xm >> 32;
135 lym = ym;
136 hym = ym >> 32;
137
138 lrm = DPXMULT(lxm, lym);
139 hrm = DPXMULT(hxm, hym);
140
141 t = DPXMULT(lxm, hym);
142
143 at = lrm + (t << 32);
144 hrm += at < lrm;
145 lrm = at;
146
147 hrm = hrm + (t >> 32);
148
149 t = DPXMULT(hxm, lym);
150
151 at = lrm + (t << 32);
152 hrm += at < lrm;
153 lrm = at;
154
155 hrm = hrm + (t >> 32);
156
157 rm = hrm | (lrm != 0);
158
159 /*
160 * Sticky shift down to normal rounding precision.
161 */
162 if ((s64) rm < 0) {
163 rm = (rm >> (64 - (DP_FBITS + 1 + 3))) |
164 ((rm << (DP_FBITS + 1 + 3)) != 0);
165 re++;
166 } else {
167 rm = (rm >> (64 - (DP_FBITS + 1 + 3 + 1))) |
168 ((rm << (DP_FBITS + 1 + 3 + 1)) != 0);
169 }
170 assert(rm & (DP_HIDDEN_BIT << 3));
171
172 return ieee754dp_format(rs, re, rm);
173}