Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* IEEE754 floating point arithmetic |
| 2 | * double precision: common utilities |
| 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 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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., |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 19 | * 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 20 | */ |
| 21 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | #include "ieee754dp.h" |
| 23 | |
Ralf Baechle | 2209bcb | 2014-04-16 01:31:11 +0200 | [diff] [blame] | 24 | union ieee754dp ieee754dp_add(union ieee754dp x, union ieee754dp y) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 25 | { |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 26 | int s; |
| 27 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 28 | COMPXDP; |
| 29 | COMPYDP; |
| 30 | |
| 31 | EXPLODEXDP; |
| 32 | EXPLODEYDP; |
| 33 | |
Ralf Baechle | 9e8bad1 | 2014-04-19 00:36:32 +0200 | [diff] [blame] | 34 | ieee754_clearcx(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 35 | |
| 36 | FLUSHXDP; |
| 37 | FLUSHYDP; |
| 38 | |
| 39 | switch (CLPAIR(xc, yc)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 40 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_SNAN): |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 41 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_SNAN): |
| 42 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_SNAN): |
| 43 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_SNAN): |
| 44 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_SNAN): |
Maciej W. Rozycki | d5afa7e | 2015-04-03 23:25:34 +0100 | [diff] [blame] | 45 | return ieee754dp_nanxcpt(y); |
| 46 | |
| 47 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_SNAN): |
| 48 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_QNAN): |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 49 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_ZERO): |
| 50 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_NORM): |
| 51 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_DNORM): |
| 52 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_INF): |
Maciej W. Rozycki | d5afa7e | 2015-04-03 23:25:34 +0100 | [diff] [blame] | 53 | return ieee754dp_nanxcpt(x); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 54 | |
| 55 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_QNAN): |
| 56 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_QNAN): |
| 57 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_QNAN): |
| 58 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_QNAN): |
| 59 | return y; |
| 60 | |
| 61 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_QNAN): |
| 62 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_ZERO): |
| 63 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_NORM): |
| 64 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_DNORM): |
| 65 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_INF): |
| 66 | return x; |
| 67 | |
| 68 | |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 69 | /* |
| 70 | * Infinity handling |
| 71 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 72 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF): |
| 73 | if (xs == ys) |
| 74 | return x; |
Ralf Baechle | 9e8bad1 | 2014-04-19 00:36:32 +0200 | [diff] [blame] | 75 | ieee754_setcx(IEEE754_INVALID_OPERATION); |
Ralf Baechle | 90efba3 | 2014-04-25 03:19:57 +0200 | [diff] [blame] | 76 | return ieee754dp_indef(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 77 | |
| 78 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_INF): |
| 79 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_INF): |
| 80 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_INF): |
| 81 | return y; |
| 82 | |
| 83 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_ZERO): |
| 84 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM): |
| 85 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM): |
| 86 | return x; |
| 87 | |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 88 | /* |
| 89 | * Zero handling |
| 90 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 91 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO): |
| 92 | if (xs == ys) |
| 93 | return x; |
| 94 | else |
Ralf Baechle | 56a6473 | 2014-04-30 11:21:55 +0200 | [diff] [blame] | 95 | return ieee754dp_zero(ieee754_csr.rm == FPU_CSR_RD); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 96 | |
| 97 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_ZERO): |
| 98 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_ZERO): |
| 99 | return x; |
| 100 | |
| 101 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM): |
| 102 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_DNORM): |
| 103 | return y; |
| 104 | |
| 105 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM): |
| 106 | DPDNORMX; |
| 107 | |
| 108 | /* FALL THROUGH */ |
| 109 | |
| 110 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM): |
| 111 | DPDNORMY; |
| 112 | break; |
| 113 | |
| 114 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM): |
| 115 | DPDNORMX; |
| 116 | break; |
| 117 | |
| 118 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM): |
| 119 | break; |
| 120 | } |
| 121 | assert(xm & DP_HIDDEN_BIT); |
| 122 | assert(ym & DP_HIDDEN_BIT); |
| 123 | |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 124 | /* |
| 125 | * Provide guard,round and stick bit space. |
| 126 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 127 | xm <<= 3; |
| 128 | ym <<= 3; |
| 129 | |
| 130 | if (xe > ye) { |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 131 | /* |
| 132 | * Have to shift y fraction right to align. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 133 | */ |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 134 | s = xe - ye; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 135 | ym = XDPSRS(ym, s); |
| 136 | ye += s; |
| 137 | } else if (ye > xe) { |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 138 | /* |
| 139 | * Have to shift x fraction right to align. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 140 | */ |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 141 | s = ye - xe; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 142 | xm = XDPSRS(xm, s); |
| 143 | xe += s; |
| 144 | } |
| 145 | assert(xe == ye); |
| 146 | assert(xe <= DP_EMAX); |
| 147 | |
| 148 | if (xs == ys) { |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 149 | /* |
| 150 | * Generate 28 bit result of adding two 27 bit numbers |
| 151 | * leaving result in xm, xs and xe. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 152 | */ |
| 153 | xm = xm + ym; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 154 | |
Ralf Baechle | ad8fb55 | 2014-04-22 15:51:55 +0200 | [diff] [blame] | 155 | if (xm >> (DP_FBITS + 1 + 3)) { /* carry out */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 156 | xm = XDPSRS1(xm); |
| 157 | xe++; |
| 158 | } |
| 159 | } else { |
| 160 | if (xm >= ym) { |
| 161 | xm = xm - ym; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 162 | } else { |
| 163 | xm = ym - xm; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 164 | xs = ys; |
| 165 | } |
| 166 | if (xm == 0) |
Ralf Baechle | 56a6473 | 2014-04-30 11:21:55 +0200 | [diff] [blame] | 167 | return ieee754dp_zero(ieee754_csr.rm == FPU_CSR_RD); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 168 | |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 169 | /* |
| 170 | * Normalize to rounding precision. |
| 171 | */ |
Ralf Baechle | ad8fb55 | 2014-04-22 15:51:55 +0200 | [diff] [blame] | 172 | while ((xm >> (DP_FBITS + 3)) == 0) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 173 | xm <<= 1; |
| 174 | xe--; |
| 175 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 176 | } |
Ralf Baechle | 90efba3 | 2014-04-25 03:19:57 +0200 | [diff] [blame] | 177 | |
| 178 | return ieee754dp_format(xs, xe, xm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 179 | } |