mangle.cpp

Go to the documentation of this file.

/*
 * Vega Strike
 * Copyright (C) 2002 David Jeffery
 *
 * http://vegastrike.sourceforge.net/
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version 2
 * of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
 */

/* NOTES: this documentation on this code was written on 02/04/29.  Hopefully it'll
 * be kept up to date.
 *
 * This evil code encodes floats into smaller bit lengths and decodes it back to a
 * normal float.  There are three variables that are used to determine how to encode
 * and decode the number.
 *
 * exponent:  exponent is the number of bits to use to encode the exponent portion of
 *            the floating point number.
 *
 * significand:  significand is the number of bits to use to store the significand of
 *               the floating point number.
 *
 * denorm: denorm is a evil little trick.  It lets you control at what exponent value
 *         to start storing the number in denormalized form.  Making this number
 *         positive enables you to reduce accuracy of small numbers in order to
 *         increase the range possible for the stored number.
 *
 * denorm can allow you to "cheat" and get more accuracy at large numbers by
 * increasing denorm, reducing exponent, and increasing significand.  This reduces
 * accuracy of small numbers, but you probably don't need most of the small number
 * accuracy anyway
 */

#include "mangle.h"

union type_conv
{
    float fl_val;
    unsigned int u32_val;
};

unsigned int float_to_uint32( float value, char exponent, char significand, char denorm )
{
    union type_conv tmp;
    unsigned int    newval, tmpsig;
    int tmpexp;

    tmp.fl_val = value;
    //set the sign bit
    newval     = SIGN_MASK&tmp.u32_val;
    newval   >>= EXP_BITS+SIGNIF_BITS-(exponent+significand);
    //get the exponent and remove its offset
    tmpexp     = (EXP_MASK&tmp.u32_val)>>SIGNIF_BITS;
    tmpexp    -= 127;
    //get the significand but leave one extra bit for use while performing
    //rounding later
    tmpsig     = (SIGNIF_MASK&tmp.u32_val)>>(SIGNIF_BITS-significand-1);
    //if this will be a normalized number, perform rounding
    if (tmpexp > denorm && tmpsig&1) {
        tmpsig++;
        if ( tmpsig&( 1<<(significand+1) ) ) {
            tmpexp++;
            tmpsig -= 1<<(significand+1);
        }
    }
    //remove the bit saved for rounding
    tmpsig >>= 1;
    if (tmpexp == denorm) {
        if (tmpsig&1)          //more rounding fun
            tmpsig++;
        if ( tmpsig&(1<<significand) ) {
            tmpexp++;
            tmpsig -= 1<<significand;
            tmpexp  = 1;
        } else {
            tmpsig  |= 1<<significand;
            tmpsig >>= 1;
            tmpexp   = 0;
        }
    } else if (tmpexp < denorm) {
        tmpsig  |= 1<<significand;
        tmpsig >>= denorm-tmpexp;
        if (tmpsig&1)
            tmpsig++;
        tmpsig >>= 1;
        tmpexp   = 0;
        //if the number is too big, clamp it to the maximum represented value
    } else if (tmpexp >= (1<<exponent)+denorm) {
        tmpexp = BITMASK( exponent );
        tmpsig = BITMASK( significand );
    } else {
        tmpexp -= denorm;
    }
    newval |= (tmpexp<<significand)|tmpsig;
    return newval;
}

float uint32_to_float( unsigned int value, char exponent, char significand, char denorm )
{
    union type_conv newval;
    unsigned int    tmpsig;
    int tmpexp;

    newval.u32_val   = ( 1<<(exponent+significand) )&value;
    newval.u32_val <<= 31-(exponent+significand);
    tmpsig   = value&BITMASK( significand );
    tmpexp   = value&(BITMASK( exponent )<<significand);
    tmpexp >>= significand;
    //if the number should be 0 or -0, go ahead and return
    if (tmpexp == 0 && tmpsig == 0) {
        return newval.fl_val;
        //if a denormalized value, rebuild to normalized form
    } else if (tmpexp == 0) {
        tmpsig <<= 1;
        while ( !( tmpsig&(1<<significand) ) ) {
            tmpexp  -= 1;
            tmpsig <<= 1;
        }
        tmpsig -= 1<<significand;
    }
    tmpexp  += 127+denorm;
    tmpsig <<= SIGNIF_BITS-significand;

    newval.u32_val |= tmpexp<<SIGNIF_BITS;
    newval.u32_val |= tmpsig;

    return newval.fl_val;
}