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_sub(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)) { |
| 40 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_QNAN): |
| 41 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_SNAN): |
| 42 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_SNAN): |
| 43 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_SNAN): |
| 44 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_SNAN): |
| 45 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_SNAN): |
| 46 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_SNAN): |
| 47 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_ZERO): |
| 48 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_NORM): |
| 49 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_DNORM): |
| 50 | case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_INF): |
Ralf Baechle | 9e8bad1 | 2014-04-19 00:36:32 +0200 | [diff] [blame] | 51 | ieee754_setcx(IEEE754_INVALID_OPERATION); |
Ralf Baechle | 90efba3 | 2014-04-25 03:19:57 +0200 | [diff] [blame] | 52 | return ieee754dp_nanxcpt(ieee754dp_indef()); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 53 | |
| 54 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_QNAN): |
| 55 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_QNAN): |
| 56 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_QNAN): |
| 57 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_QNAN): |
| 58 | return y; |
| 59 | |
| 60 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_QNAN): |
| 61 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_ZERO): |
| 62 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_NORM): |
| 63 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_DNORM): |
| 64 | case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_INF): |
| 65 | return x; |
| 66 | |
| 67 | |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 68 | /* |
| 69 | * Infinity handling |
| 70 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 71 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF): |
| 72 | if (xs != ys) |
| 73 | return x; |
Ralf Baechle | 9e8bad1 | 2014-04-19 00:36:32 +0200 | [diff] [blame] | 74 | ieee754_setcx(IEEE754_INVALID_OPERATION); |
Ralf Baechle | 90efba3 | 2014-04-25 03:19:57 +0200 | [diff] [blame] | 75 | return ieee754dp_indef(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 76 | |
| 77 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_INF): |
| 78 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_INF): |
| 79 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_INF): |
| 80 | return ieee754dp_inf(ys ^ 1); |
| 81 | |
| 82 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_ZERO): |
| 83 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM): |
| 84 | case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM): |
| 85 | return x; |
| 86 | |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 87 | /* |
| 88 | * Zero handling |
| 89 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 90 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO): |
| 91 | if (xs != ys) |
| 92 | return x; |
| 93 | else |
Ralf Baechle | 56a6473 | 2014-04-30 11:21:55 +0200 | [diff] [blame^] | 94 | return ieee754dp_zero(ieee754_csr.rm == FPU_CSR_RD); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 95 | |
| 96 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_ZERO): |
| 97 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_ZERO): |
| 98 | return x; |
| 99 | |
| 100 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM): |
| 101 | case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_DNORM): |
| 102 | /* quick fix up */ |
| 103 | DPSIGN(y) ^= 1; |
| 104 | return y; |
| 105 | |
| 106 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM): |
| 107 | DPDNORMX; |
André Goddard Rosa | af901ca | 2009-11-14 13:09:05 -0200 | [diff] [blame] | 108 | /* FALL THROUGH */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 109 | |
| 110 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM): |
| 111 | /* normalize ym,ye */ |
| 112 | DPDNORMY; |
| 113 | break; |
| 114 | |
| 115 | case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM): |
| 116 | /* normalize xm,xe */ |
| 117 | DPDNORMX; |
| 118 | break; |
| 119 | |
| 120 | case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM): |
| 121 | break; |
| 122 | } |
| 123 | /* flip sign of y and handle as add */ |
| 124 | ys ^= 1; |
| 125 | |
| 126 | assert(xm & DP_HIDDEN_BIT); |
| 127 | assert(ym & DP_HIDDEN_BIT); |
| 128 | |
| 129 | |
| 130 | /* provide guard,round and stick bit dpace */ |
| 131 | xm <<= 3; |
| 132 | ym <<= 3; |
| 133 | |
| 134 | if (xe > ye) { |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 135 | /* |
| 136 | * Have to shift y fraction right to align |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 137 | */ |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 138 | s = xe - ye; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 139 | ym = XDPSRS(ym, s); |
| 140 | ye += s; |
| 141 | } else if (ye > xe) { |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 142 | /* |
| 143 | * Have to shift x fraction right to align |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 144 | */ |
Ralf Baechle | 3f7cac4 | 2014-04-26 01:49:14 +0200 | [diff] [blame] | 145 | s = ye - xe; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 146 | xm = XDPSRS(xm, s); |
| 147 | xe += s; |
| 148 | } |
| 149 | assert(xe == ye); |
| 150 | assert(xe <= DP_EMAX); |
| 151 | |
| 152 | if (xs == ys) { |
| 153 | /* generate 28 bit result of adding two 27 bit numbers |
| 154 | */ |
| 155 | xm = xm + ym; |
| 156 | xe = xe; |
| 157 | xs = xs; |
| 158 | |
Ralf Baechle | ad8fb55 | 2014-04-22 15:51:55 +0200 | [diff] [blame] | 159 | if (xm >> (DP_FBITS + 1 + 3)) { /* carry out */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 160 | xm = XDPSRS1(xm); /* shift preserving sticky */ |
| 161 | xe++; |
| 162 | } |
| 163 | } else { |
| 164 | if (xm >= ym) { |
| 165 | xm = xm - ym; |
| 166 | xe = xe; |
| 167 | xs = xs; |
| 168 | } else { |
| 169 | xm = ym - xm; |
| 170 | xe = xe; |
| 171 | xs = ys; |
| 172 | } |
| 173 | if (xm == 0) { |
Ralf Baechle | 56a6473 | 2014-04-30 11:21:55 +0200 | [diff] [blame^] | 174 | if (ieee754_csr.rm == FPU_CSR_RD) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 175 | return ieee754dp_zero(1); /* round negative inf. => sign = -1 */ |
| 176 | else |
| 177 | return ieee754dp_zero(0); /* other round modes => sign = 1 */ |
| 178 | } |
| 179 | |
| 180 | /* normalize to rounding precision |
| 181 | */ |
Ralf Baechle | ad8fb55 | 2014-04-22 15:51:55 +0200 | [diff] [blame] | 182 | while ((xm >> (DP_FBITS + 3)) == 0) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 183 | xm <<= 1; |
| 184 | xe--; |
| 185 | } |
| 186 | } |
Ralf Baechle | 90efba3 | 2014-04-25 03:19:57 +0200 | [diff] [blame] | 187 | |
| 188 | return ieee754dp_format(xs, xe, xm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 189 | } |