Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* IEEE754 floating point arithmetic |
| 2 | * single precision |
| 3 | */ |
| 4 | /* |
| 5 | * MIPS floating point support |
| 6 | * Copyright (C) 1994-2000 Algorithmics Ltd. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | * |
| 8 | * ######################################################################## |
| 9 | * |
| 10 | * This program is free software; you can distribute it and/or modify it |
| 11 | * under the terms of the GNU General Public License (Version 2) as |
| 12 | * published by the Free Software Foundation. |
| 13 | * |
| 14 | * This program is distributed in the hope it will be useful, but WITHOUT |
| 15 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
| 16 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
| 17 | * for more details. |
| 18 | * |
| 19 | * You should have received a copy of the GNU General Public License along |
| 20 | * with this program; if not, write to the Free Software Foundation, Inc., |
| 21 | * 59 Temple Place - Suite 330, Boston MA 02111-1307, USA. |
| 22 | * |
| 23 | * ######################################################################## |
| 24 | */ |
| 25 | |
| 26 | |
| 27 | #include "ieee754sp.h" |
| 28 | |
| 29 | ieee754sp ieee754sp_mul(ieee754sp x, ieee754sp y) |
| 30 | { |
| 31 | COMPXSP; |
| 32 | COMPYSP; |
| 33 | |
| 34 | EXPLODEXSP; |
| 35 | EXPLODEYSP; |
| 36 | |
| 37 | CLEARCX; |
| 38 | |
| 39 | FLUSHXSP; |
| 40 | FLUSHYSP; |
| 41 | |
| 42 | switch (CLPAIR(xc, yc)) { |
| 43 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_QNAN): |
| 44 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_SNAN): |
| 45 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_SNAN): |
| 46 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_SNAN): |
| 47 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_SNAN): |
| 48 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_SNAN): |
| 49 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_SNAN): |
| 50 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_ZERO): |
| 51 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_NORM): |
| 52 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_DNORM): |
| 53 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_INF): |
| 54 | SETCX(IEEE754_INVALID_OPERATION); |
| 55 | return ieee754sp_nanxcpt(ieee754sp_indef(), "mul", x, y); |
| 56 | |
| 57 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_QNAN): |
| 58 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_QNAN): |
| 59 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_QNAN): |
| 60 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_QNAN): |
| 61 | return y; |
| 62 | |
| 63 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_QNAN): |
| 64 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_ZERO): |
| 65 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_NORM): |
| 66 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_DNORM): |
| 67 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_INF): |
| 68 | return x; |
| 69 | |
| 70 | |
| 71 | /* Infinity handling */ |
| 72 | |
| 73 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_ZERO): |
| 74 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_INF): |
| 75 | SETCX(IEEE754_INVALID_OPERATION); |
| 76 | return ieee754sp_xcpt(ieee754sp_indef(), "mul", x, y); |
| 77 | |
| 78 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_INF): |
| 79 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_INF): |
| 80 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM): |
| 81 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM): |
| 82 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF): |
| 83 | return ieee754sp_inf(xs ^ ys); |
| 84 | |
| 85 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO): |
| 86 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM): |
| 87 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_DNORM): |
| 88 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_ZERO): |
| 89 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_ZERO): |
| 90 | return ieee754sp_zero(xs ^ ys); |
| 91 | |
| 92 | |
| 93 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM): |
| 94 | SPDNORMX; |
| 95 | |
| 96 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM): |
| 97 | SPDNORMY; |
| 98 | break; |
| 99 | |
| 100 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM): |
| 101 | SPDNORMX; |
| 102 | break; |
| 103 | |
| 104 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM): |
| 105 | break; |
| 106 | } |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame^] | 107 | /* rm = xm * ym, re = xe+ye basically */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 108 | assert(xm & SP_HIDDEN_BIT); |
| 109 | assert(ym & SP_HIDDEN_BIT); |
| 110 | |
| 111 | { |
| 112 | int re = xe + ye; |
| 113 | int rs = xs ^ ys; |
| 114 | unsigned rm; |
| 115 | |
| 116 | /* shunt to top of word */ |
| 117 | xm <<= 32 - (SP_MBITS + 1); |
| 118 | ym <<= 32 - (SP_MBITS + 1); |
| 119 | |
| 120 | /* multiply 32bits xm,ym to give high 32bits rm with stickness |
| 121 | */ |
| 122 | { |
| 123 | unsigned short lxm = xm & 0xffff; |
| 124 | unsigned short hxm = xm >> 16; |
| 125 | unsigned short lym = ym & 0xffff; |
| 126 | unsigned short hym = ym >> 16; |
| 127 | unsigned lrm; |
| 128 | unsigned hrm; |
| 129 | |
| 130 | lrm = lxm * lym; /* 16 * 16 => 32 */ |
| 131 | hrm = hxm * hym; /* 16 * 16 => 32 */ |
| 132 | |
| 133 | { |
| 134 | unsigned t = lxm * hym; /* 16 * 16 => 32 */ |
| 135 | { |
| 136 | unsigned at = lrm + (t << 16); |
| 137 | hrm += at < lrm; |
| 138 | lrm = at; |
| 139 | } |
| 140 | hrm = hrm + (t >> 16); |
| 141 | } |
| 142 | |
| 143 | { |
| 144 | unsigned t = hxm * lym; /* 16 * 16 => 32 */ |
| 145 | { |
| 146 | unsigned at = lrm + (t << 16); |
| 147 | hrm += at < lrm; |
| 148 | lrm = at; |
| 149 | } |
| 150 | hrm = hrm + (t >> 16); |
| 151 | } |
| 152 | rm = hrm | (lrm != 0); |
| 153 | } |
| 154 | |
| 155 | /* |
| 156 | * sticky shift down to normal rounding precision |
| 157 | */ |
| 158 | if ((int) rm < 0) { |
| 159 | rm = (rm >> (32 - (SP_MBITS + 1 + 3))) | |
| 160 | ((rm << (SP_MBITS + 1 + 3)) != 0); |
| 161 | re++; |
| 162 | } else { |
| 163 | rm = (rm >> (32 - (SP_MBITS + 1 + 3 + 1))) | |
| 164 | ((rm << (SP_MBITS + 1 + 3 + 1)) != 0); |
| 165 | } |
| 166 | assert(rm & (SP_HIDDEN_BIT << 3)); |
| 167 | |
| 168 | SPNORMRET2(rs, re, rm, "mul", x, y); |
| 169 | } |
| 170 | } |