Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Linux/PA-RISC Project (http://www.parisc-linux.org/) |
| 3 | * |
| 4 | * Floating-point emulation code |
| 5 | * Copyright (C) 2001 Hewlett-Packard (Paul Bame) <bame@debian.org> |
| 6 | * |
| 7 | * This program is free software; you can redistribute it and/or modify |
| 8 | * it under the terms of the GNU General Public License as published by |
| 9 | * the Free Software Foundation; either version 2, or (at your option) |
| 10 | * any later version. |
| 11 | * |
| 12 | * This program is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 15 | * GNU General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU General Public License |
| 18 | * along with this program; if not, write to the Free Software |
| 19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
| 20 | */ |
| 21 | /* |
| 22 | * BEGIN_DESC |
| 23 | * |
| 24 | * File: |
| 25 | * @(#) pa/spmath/sfsqrt.c $Revision: 1.1 $ |
| 26 | * |
| 27 | * Purpose: |
| 28 | * Single Floating-point Square Root |
| 29 | * |
| 30 | * External Interfaces: |
| 31 | * sgl_fsqrt(srcptr,nullptr,dstptr,status) |
| 32 | * |
| 33 | * Internal Interfaces: |
| 34 | * |
| 35 | * Theory: |
| 36 | * <<please update with a overview of the operation of this file>> |
| 37 | * |
| 38 | * END_DESC |
| 39 | */ |
| 40 | |
| 41 | |
| 42 | #include "float.h" |
| 43 | #include "sgl_float.h" |
| 44 | |
| 45 | /* |
| 46 | * Single Floating-point Square Root |
| 47 | */ |
| 48 | |
| 49 | /*ARGSUSED*/ |
| 50 | unsigned int |
| 51 | sgl_fsqrt( |
| 52 | sgl_floating_point *srcptr, |
| 53 | unsigned int *nullptr, |
| 54 | sgl_floating_point *dstptr, |
| 55 | unsigned int *status) |
| 56 | { |
| 57 | register unsigned int src, result; |
| 58 | register int src_exponent; |
| 59 | register unsigned int newbit, sum; |
| 60 | register boolean guardbit = FALSE, even_exponent; |
| 61 | |
| 62 | src = *srcptr; |
| 63 | /* |
| 64 | * check source operand for NaN or infinity |
| 65 | */ |
| 66 | if ((src_exponent = Sgl_exponent(src)) == SGL_INFINITY_EXPONENT) { |
| 67 | /* |
| 68 | * is signaling NaN? |
| 69 | */ |
| 70 | if (Sgl_isone_signaling(src)) { |
| 71 | /* trap if INVALIDTRAP enabled */ |
| 72 | if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION); |
| 73 | /* make NaN quiet */ |
| 74 | Set_invalidflag(); |
| 75 | Sgl_set_quiet(src); |
| 76 | } |
| 77 | /* |
| 78 | * Return quiet NaN or positive infinity. |
Simon Arlott | 7022672 | 2007-05-11 20:42:34 +0100 | [diff] [blame] | 79 | * Fall through to negative test if negative infinity. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 80 | */ |
| 81 | if (Sgl_iszero_sign(src) || Sgl_isnotzero_mantissa(src)) { |
| 82 | *dstptr = src; |
| 83 | return(NOEXCEPTION); |
| 84 | } |
| 85 | } |
| 86 | |
| 87 | /* |
| 88 | * check for zero source operand |
| 89 | */ |
| 90 | if (Sgl_iszero_exponentmantissa(src)) { |
| 91 | *dstptr = src; |
| 92 | return(NOEXCEPTION); |
| 93 | } |
| 94 | |
| 95 | /* |
| 96 | * check for negative source operand |
| 97 | */ |
| 98 | if (Sgl_isone_sign(src)) { |
| 99 | /* trap if INVALIDTRAP enabled */ |
| 100 | if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION); |
| 101 | /* make NaN quiet */ |
| 102 | Set_invalidflag(); |
| 103 | Sgl_makequietnan(src); |
| 104 | *dstptr = src; |
| 105 | return(NOEXCEPTION); |
| 106 | } |
| 107 | |
| 108 | /* |
| 109 | * Generate result |
| 110 | */ |
| 111 | if (src_exponent > 0) { |
| 112 | even_exponent = Sgl_hidden(src); |
| 113 | Sgl_clear_signexponent_set_hidden(src); |
| 114 | } |
| 115 | else { |
| 116 | /* normalize operand */ |
| 117 | Sgl_clear_signexponent(src); |
| 118 | src_exponent++; |
| 119 | Sgl_normalize(src,src_exponent); |
| 120 | even_exponent = src_exponent & 1; |
| 121 | } |
| 122 | if (even_exponent) { |
| 123 | /* exponent is even */ |
| 124 | /* Add comment here. Explain why odd exponent needs correction */ |
| 125 | Sgl_leftshiftby1(src); |
| 126 | } |
| 127 | /* |
| 128 | * Add comment here. Explain following algorithm. |
| 129 | * |
| 130 | * Trust me, it works. |
| 131 | * |
| 132 | */ |
| 133 | Sgl_setzero(result); |
| 134 | newbit = 1 << SGL_P; |
| 135 | while (newbit && Sgl_isnotzero(src)) { |
| 136 | Sgl_addition(result,newbit,sum); |
| 137 | if(sum <= Sgl_all(src)) { |
| 138 | /* update result */ |
| 139 | Sgl_addition(result,(newbit<<1),result); |
| 140 | Sgl_subtract(src,sum,src); |
| 141 | } |
| 142 | Sgl_rightshiftby1(newbit); |
| 143 | Sgl_leftshiftby1(src); |
| 144 | } |
| 145 | /* correct exponent for pre-shift */ |
| 146 | if (even_exponent) { |
| 147 | Sgl_rightshiftby1(result); |
| 148 | } |
| 149 | |
| 150 | /* check for inexact */ |
| 151 | if (Sgl_isnotzero(src)) { |
| 152 | if (!even_exponent && Sgl_islessthan(result,src)) |
| 153 | Sgl_increment(result); |
| 154 | guardbit = Sgl_lowmantissa(result); |
| 155 | Sgl_rightshiftby1(result); |
| 156 | |
| 157 | /* now round result */ |
| 158 | switch (Rounding_mode()) { |
| 159 | case ROUNDPLUS: |
| 160 | Sgl_increment(result); |
| 161 | break; |
| 162 | case ROUNDNEAREST: |
| 163 | /* stickybit is always true, so guardbit |
| 164 | * is enough to determine rounding */ |
| 165 | if (guardbit) { |
| 166 | Sgl_increment(result); |
| 167 | } |
| 168 | break; |
| 169 | } |
| 170 | /* increment result exponent by 1 if mantissa overflowed */ |
| 171 | if (Sgl_isone_hiddenoverflow(result)) src_exponent+=2; |
| 172 | |
| 173 | if (Is_inexacttrap_enabled()) { |
| 174 | Sgl_set_exponent(result, |
| 175 | ((src_exponent-SGL_BIAS)>>1)+SGL_BIAS); |
| 176 | *dstptr = result; |
| 177 | return(INEXACTEXCEPTION); |
| 178 | } |
| 179 | else Set_inexactflag(); |
| 180 | } |
| 181 | else { |
| 182 | Sgl_rightshiftby1(result); |
| 183 | } |
| 184 | Sgl_set_exponent(result,((src_exponent-SGL_BIAS)>>1)+SGL_BIAS); |
| 185 | *dstptr = result; |
| 186 | return(NOEXCEPTION); |
| 187 | } |