/**CFile***********************************************************************

  FileName    [epd.c]

  PackageName [epd]

  Synopsis    [Arithmetic functions with extended double precision.]

  Description []

  SeeAlso     []

  Author      [In-Ho Moon]

  Copyright   [Copyright (c) 1995-2004, Regents of the University of Colorado

  All rights reserved.

  Redistribution and use in source and binary forms, with or without
  modification, are permitted provided that the following conditions
  are met:

  Redistributions of source code must retain the above copyright
  notice, this list of conditions and the following disclaimer.

  Redistributions in binary form must reproduce the above copyright
  notice, this list of conditions and the following disclaimer in the
  documentation and/or other materials provided with the distribution.

  Neither the name of the University of Colorado nor the names of its
  contributors may be used to endorse or promote products derived from
  this software without specific prior written permission.

  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  POSSIBILITY OF SUCH DAMAGE.]

  Revision    [$Id: epd.c,v 1.10 2004/08/13 18:20:30 fabio Exp $]

******************************************************************************/

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include "misc/util/util_hack.h"
#include "epd.h"

ABC_NAMESPACE_IMPL_START

/**Function********************************************************************

  Synopsis    [Allocates an EpDouble struct.]

  Description [Allocates an EpDouble struct.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
EpDouble *
EpdAlloc(void)
{
  EpDouble      *epd;

  epd = ABC_ALLOC(EpDouble, 1);
  return(epd);
}


/**Function********************************************************************

  Synopsis    [Compares two EpDouble struct.]

  Description [Compares two EpDouble struct.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
EpdCmp(const char *key1, const char *key2)
{
  EpDouble *epd1 = (EpDouble *) key1;
  EpDouble *epd2 = (EpDouble *) key2;
  if (epd1->type.value != epd2->type.value ||
      epd1->exponent != epd2->exponent) {
    return(1);
  }
  return(0);
}


/**Function********************************************************************

  Synopsis    [Frees an EpDouble struct.]

  Description [Frees an EpDouble struct.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdFree(EpDouble *epd)
{
  ABC_FREE(epd);
}


/**Function********************************************************************

  Synopsis    [Converts an arbitrary precision double value to a string.]

  Description [Converts an arbitrary precision double value to a string.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdGetString(EpDouble *epd, char *str)
{
  double        value;
  int           exponent;
  char          *pos;

  if (IsNanDouble(epd->type.value)) {
    sprintf(str, "NaN");
    return;
  } else if (IsInfDouble(epd->type.value)) {
    if (epd->type.bits.sign == 1)
      sprintf(str, "-Inf");
    else
      sprintf(str, "Inf");
    return;
  }

  assert(epd->type.bits.exponent == EPD_MAX_BIN ||
         epd->type.bits.exponent == 0);

  EpdGetValueAndDecimalExponent(epd, &value, &exponent);
  sprintf(str, "%e", value);
  pos = strstr(str, "e");
  if (exponent >= 0) {
    if (exponent < 10)
      sprintf(pos + 1, "+0%d", exponent);
    else
      sprintf(pos + 1, "+%d", exponent);
  } else {
    exponent *= -1;
    if (exponent < 10)
      sprintf(pos + 1, "-0%d", exponent);
    else
      sprintf(pos + 1, "-%d", exponent);
  }
}


/**Function********************************************************************

  Synopsis    [Converts double to EpDouble struct.]

  Description [Converts double to EpDouble struct.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdConvert(double value, EpDouble *epd)
{
  epd->type.value = value;
  epd->exponent = 0;
  EpdNormalize(epd);
}


/**Function********************************************************************

  Synopsis    [Multiplies two arbitrary precision double values.]

  Description [Multiplies two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdMultiply(EpDouble *epd1, double value)
{
  EpDouble      epd2;
  double        tmp;
  int           exponent;

  if (EpdIsNan(epd1) || IsNanDouble(value)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || IsInfDouble(value)) {
    int sign;

    EpdConvert(value, &epd2);
    sign = epd1->type.bits.sign ^ epd2.type.bits.sign;
    EpdMakeInf(epd1, sign);
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);

  EpdConvert(value, &epd2);
  tmp = epd1->type.value * epd2.type.value;
  exponent = epd1->exponent + epd2.exponent;
  epd1->type.value = tmp;
  epd1->exponent = exponent;
  EpdNormalize(epd1);
}


/**Function********************************************************************

  Synopsis    [Multiplies two arbitrary precision double values.]

  Description [Multiplies two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdMultiply2(EpDouble *epd1, EpDouble *epd2)
{
  double        value;
  int           exponent;

  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
    EpdMakeInf(epd1, sign);
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);
  assert(epd2->type.bits.exponent == EPD_MAX_BIN);

  value = epd1->type.value * epd2->type.value;
  exponent = epd1->exponent + epd2->exponent;
  epd1->type.value = value;
  epd1->exponent = exponent;
  EpdNormalize(epd1);
}


/**Function********************************************************************

  Synopsis    [Multiplies two arbitrary precision double values.]

  Description [Multiplies two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdMultiply2Decimal(EpDouble *epd1, EpDouble *epd2)
{
  double        value;
  int           exponent;

  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
    EpdMakeInf(epd1, sign);
    return;
  }

  value = epd1->type.value * epd2->type.value;
  exponent = epd1->exponent + epd2->exponent;
  epd1->type.value = value;
  epd1->exponent = exponent;
  EpdNormalizeDecimal(epd1);
}


/**Function********************************************************************

  Synopsis    [Multiplies two arbitrary precision double values.]

  Description [Multiplies two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdMultiply3(EpDouble *epd1, EpDouble *epd2, EpDouble *epd3)
{
  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
    EpdMakeInf(epd3, sign);
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);
  assert(epd2->type.bits.exponent == EPD_MAX_BIN);

  epd3->type.value = epd1->type.value * epd2->type.value;
  epd3->exponent = epd1->exponent + epd2->exponent;
  EpdNormalize(epd3);
}


/**Function********************************************************************

  Synopsis    [Multiplies two arbitrary precision double values.]

  Description [Multiplies two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdMultiply3Decimal(EpDouble *epd1, EpDouble *epd2, EpDouble *epd3)
{
  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
    EpdMakeInf(epd3, sign);
    return;
  }

  epd3->type.value = epd1->type.value * epd2->type.value;
  epd3->exponent = epd1->exponent + epd2->exponent;
  EpdNormalizeDecimal(epd3);
}


/**Function********************************************************************

  Synopsis    [Divides two arbitrary precision double values.]

  Description [Divides two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdDivide(EpDouble *epd1, double value)
{
  EpDouble      epd2;
  double        tmp;
  int           exponent;

  if (EpdIsNan(epd1) || IsNanDouble(value)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || IsInfDouble(value)) {
    int sign;

    EpdConvert(value, &epd2);
    if (EpdIsInf(epd1) && IsInfDouble(value)) {
      EpdMakeNan(epd1);
    } else if (EpdIsInf(epd1)) {
      sign = epd1->type.bits.sign ^ epd2.type.bits.sign;
      EpdMakeInf(epd1, sign);
    } else {
      sign = epd1->type.bits.sign ^ epd2.type.bits.sign;
      EpdMakeZero(epd1, sign);
    }
    return;
  }

  if (value == 0.0) {
    EpdMakeNan(epd1);
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);

  EpdConvert(value, &epd2);
  tmp = epd1->type.value / epd2.type.value;
  exponent = epd1->exponent - epd2.exponent;
  epd1->type.value = tmp;
  epd1->exponent = exponent;
  EpdNormalize(epd1);
}


/**Function********************************************************************

  Synopsis    [Divides two arbitrary precision double values.]

  Description [Divides two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdDivide2(EpDouble *epd1, EpDouble *epd2)
{
  double        value;
  int           exponent;

  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    if (EpdIsInf(epd1) && EpdIsInf(epd2)) {
      EpdMakeNan(epd1);
    } else if (EpdIsInf(epd1)) {
      sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
      EpdMakeInf(epd1, sign);
    } else {
      sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
      EpdMakeZero(epd1, sign);
    }
    return;
  }

  if (epd2->type.value == 0.0) {
    EpdMakeNan(epd1);
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);
  assert(epd2->type.bits.exponent == EPD_MAX_BIN);

  value = epd1->type.value / epd2->type.value;
  exponent = epd1->exponent - epd2->exponent;
  epd1->type.value = value;
  epd1->exponent = exponent;
  EpdNormalize(epd1);
}


/**Function********************************************************************

  Synopsis    [Divides two arbitrary precision double values.]

  Description [Divides two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdDivide3(EpDouble *epd1, EpDouble *epd2, EpDouble *epd3)
{
  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd3);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    if (EpdIsInf(epd1) && EpdIsInf(epd2)) {
      EpdMakeNan(epd3);
    } else if (EpdIsInf(epd1)) {
      sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
      EpdMakeInf(epd3, sign);
    } else {
      sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
      EpdMakeZero(epd3, sign);
    }
    return;
  }

  if (epd2->type.value == 0.0) {
    EpdMakeNan(epd3);
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);
  assert(epd2->type.bits.exponent == EPD_MAX_BIN);

  epd3->type.value = epd1->type.value / epd2->type.value;
  epd3->exponent = epd1->exponent - epd2->exponent;
  EpdNormalize(epd3);
}


/**Function********************************************************************

  Synopsis    [Adds two arbitrary precision double values.]

  Description [Adds two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdAdd(EpDouble *epd1, double value)
{
  EpDouble      epd2;
  double        tmp;
  int           exponent, diff;

  if (EpdIsNan(epd1) || IsNanDouble(value)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || IsInfDouble(value)) {
    int sign;

    EpdConvert(value, &epd2);
    if (EpdIsInf(epd1) && IsInfDouble(value)) {
      sign = epd1->type.bits.sign ^ epd2.type.bits.sign;
      if (sign == 1)
        EpdMakeNan(epd1);
    } else if (EpdIsInf(&epd2)) {
      EpdCopy(&epd2, epd1);
    }
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);

  EpdConvert(value, &epd2);
  if (epd1->exponent > epd2.exponent) {
    diff = epd1->exponent - epd2.exponent;
    if (diff <= EPD_MAX_BIN)
      tmp = epd1->type.value + epd2.type.value / pow((double)2.0, (double)diff);
    else
      tmp = epd1->type.value;
    exponent = epd1->exponent;
  } else if (epd1->exponent < epd2.exponent) {
    diff = epd2.exponent - epd1->exponent;
    if (diff <= EPD_MAX_BIN)
      tmp = epd1->type.value / pow((double)2.0, (double)diff) + epd2.type.value;
    else
      tmp = epd2.type.value;
    exponent = epd2.exponent;
  } else {
    tmp = epd1->type.value + epd2.type.value;
    exponent = epd1->exponent;
  }
  epd1->type.value = tmp;
  epd1->exponent = exponent;
  EpdNormalize(epd1);
}


/**Function********************************************************************

  Synopsis    [Adds two arbitrary precision double values.]

  Description [Adds two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdAdd2(EpDouble *epd1, EpDouble *epd2)
{
  double        value;
  int           exponent, diff;

  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    if (EpdIsInf(epd1) && EpdIsInf(epd2)) {
      sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
      if (sign == 1)
        EpdMakeNan(epd1);
    } else if (EpdIsInf(epd2)) {
      EpdCopy(epd2, epd1);
    }
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);
  assert(epd2->type.bits.exponent == EPD_MAX_BIN);

  if (epd1->exponent > epd2->exponent) {
    diff = epd1->exponent - epd2->exponent;
    if (diff <= EPD_MAX_BIN) {
      value = epd1->type.value +
                epd2->type.value / pow((double)2.0, (double)diff);
    } else
      value = epd1->type.value;
    exponent = epd1->exponent;
  } else if (epd1->exponent < epd2->exponent) {
    diff = epd2->exponent - epd1->exponent;
    if (diff <= EPD_MAX_BIN) {
      value = epd1->type.value / pow((double)2.0, (double)diff) +
                epd2->type.value;
    } else
      value = epd2->type.value;
    exponent = epd2->exponent;
  } else {
    value = epd1->type.value + epd2->type.value;
    exponent = epd1->exponent;
  }
  epd1->type.value = value;
  epd1->exponent = exponent;
  EpdNormalize(epd1);
}


/**Function********************************************************************

  Synopsis    [Adds two arbitrary precision double values.]

  Description [Adds two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdAdd3(EpDouble *epd1, EpDouble *epd2, EpDouble *epd3)
{
  double        value;
  int           exponent, diff;

  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd3);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    if (EpdIsInf(epd1) && EpdIsInf(epd2)) {
      sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
      if (sign == 1)
        EpdMakeNan(epd3);
      else
        EpdCopy(epd1, epd3);
    } else if (EpdIsInf(epd1)) {
      EpdCopy(epd1, epd3);
    } else {
      EpdCopy(epd2, epd3);
    }
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);
  assert(epd2->type.bits.exponent == EPD_MAX_BIN);

  if (epd1->exponent > epd2->exponent) {
    diff = epd1->exponent - epd2->exponent;
    if (diff <= EPD_MAX_BIN) {
      value = epd1->type.value +
                epd2->type.value / pow((double)2.0, (double)diff);
    } else
      value = epd1->type.value;
    exponent = epd1->exponent;
  } else if (epd1->exponent < epd2->exponent) {
    diff = epd2->exponent - epd1->exponent;
    if (diff <= EPD_MAX_BIN) {
      value = epd1->type.value / pow((double)2.0, (double)diff) +
                epd2->type.value;
    } else
      value = epd2->type.value;
    exponent = epd2->exponent;
  } else {
    value = epd1->type.value + epd2->type.value;
    exponent = epd1->exponent;
  }
  epd3->type.value = value;
  epd3->exponent = exponent;
  EpdNormalize(epd3);
}


/**Function********************************************************************

  Synopsis    [Subtracts two arbitrary precision double values.]

  Description [Subtracts two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdSubtract(EpDouble *epd1, double value)
{
  EpDouble      epd2;
  double        tmp;
  int           exponent, diff;

  if (EpdIsNan(epd1) || IsNanDouble(value)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || IsInfDouble(value)) {
    int sign;

    EpdConvert(value, &epd2);
    if (EpdIsInf(epd1) && IsInfDouble(value)) {
      sign = epd1->type.bits.sign ^ epd2.type.bits.sign;
      if (sign == 0)
        EpdMakeNan(epd1);
    } else if (EpdIsInf(&epd2)) {
      EpdCopy(&epd2, epd1);
    }
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);

  EpdConvert(value, &epd2);
  if (epd1->exponent > epd2.exponent) {
    diff = epd1->exponent - epd2.exponent;
    if (diff <= EPD_MAX_BIN)
      tmp = epd1->type.value - epd2.type.value / pow((double)2.0, (double)diff);
    else
      tmp = epd1->type.value;
    exponent = epd1->exponent;
  } else if (epd1->exponent < epd2.exponent) {
    diff = epd2.exponent - epd1->exponent;
    if (diff <= EPD_MAX_BIN)
      tmp = epd1->type.value / pow((double)2.0, (double)diff) - epd2.type.value;
    else
      tmp = epd2.type.value * (double)(-1.0);
    exponent = epd2.exponent;
  } else {
    tmp = epd1->type.value - epd2.type.value;
    exponent = epd1->exponent;
  }
  epd1->type.value = tmp;
  epd1->exponent = exponent;
  EpdNormalize(epd1);
}


/**Function********************************************************************

  Synopsis    [Subtracts two arbitrary precision double values.]

  Description [Subtracts two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdSubtract2(EpDouble *epd1, EpDouble *epd2)
{
  double        value;
  int           exponent, diff;

  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd1);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    if (EpdIsInf(epd1) && EpdIsInf(epd2)) {
      sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
      if (sign == 0)
        EpdMakeNan(epd1);
    } else if (EpdIsInf(epd2)) {
      EpdCopy(epd2, epd1);
    }
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);
  assert(epd2->type.bits.exponent == EPD_MAX_BIN);

  if (epd1->exponent > epd2->exponent) {
    diff = epd1->exponent - epd2->exponent;
    if (diff <= EPD_MAX_BIN) {
      value = epd1->type.value -
                epd2->type.value / pow((double)2.0, (double)diff);
    } else
      value = epd1->type.value;
    exponent = epd1->exponent;
  } else if (epd1->exponent < epd2->exponent) {
    diff = epd2->exponent - epd1->exponent;
    if (diff <= EPD_MAX_BIN) {
      value = epd1->type.value / pow((double)2.0, (double)diff) -
                epd2->type.value;
    } else
      value = epd2->type.value * (double)(-1.0);
    exponent = epd2->exponent;
  } else {
    value = epd1->type.value - epd2->type.value;
    exponent = epd1->exponent;
  }
  epd1->type.value = value;
  epd1->exponent = exponent;
  EpdNormalize(epd1);
}


/**Function********************************************************************

  Synopsis    [Subtracts two arbitrary precision double values.]

  Description [Subtracts two arbitrary precision double values.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdSubtract3(EpDouble *epd1, EpDouble *epd2, EpDouble *epd3)
{
  double        value;
  int           exponent, diff;

  if (EpdIsNan(epd1) || EpdIsNan(epd2)) {
    EpdMakeNan(epd3);
    return;
  } else if (EpdIsInf(epd1) || EpdIsInf(epd2)) {
    int sign;

    if (EpdIsInf(epd1) && EpdIsInf(epd2)) {
      sign = epd1->type.bits.sign ^ epd2->type.bits.sign;
      if (sign == 0)
        EpdCopy(epd1, epd3);
      else
        EpdMakeNan(epd3);
    } else if (EpdIsInf(epd1)) {
      EpdCopy(epd1, epd1);
    } else {
      sign = epd2->type.bits.sign ^ 0x1;
      EpdMakeInf(epd3, sign);
    }
    return;
  }

  assert(epd1->type.bits.exponent == EPD_MAX_BIN);
  assert(epd2->type.bits.exponent == EPD_MAX_BIN);

  if (epd1->exponent > epd2->exponent) {
    diff = epd1->exponent - epd2->exponent;
    if (diff <= EPD_MAX_BIN) {
      value = epd1->type.value -
                epd2->type.value / pow((double)2.0, (double)diff);
    } else
      value = epd1->type.value;
    exponent = epd1->exponent;
  } else if (epd1->exponent < epd2->exponent) {
    diff = epd2->exponent - epd1->exponent;
    if (diff <= EPD_MAX_BIN) {
      value = epd1->type.value / pow((double)2.0, (double)diff) -
                epd2->type.value;
    } else
      value = epd2->type.value * (double)(-1.0);
    exponent = epd2->exponent;
  } else {
    value = epd1->type.value - epd2->type.value;
    exponent = epd1->exponent;
  }
  epd3->type.value = value;
  epd3->exponent = exponent;
  EpdNormalize(epd3);
}


/**Function********************************************************************

  Synopsis    [Computes arbitrary precision pow of base 2.]

  Description [Computes arbitrary precision pow of base 2.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdPow2(int n, EpDouble *epd)
{
  if (n <= EPD_MAX_BIN) {
    EpdConvert(pow((double)2.0, (double)n), epd);
  } else {
    EpDouble    epd1, epd2;
    int         n1, n2;

    n1 = n / 2;
    n2 = n - n1;
    EpdPow2(n1, &epd1);
    EpdPow2(n2, &epd2);
    EpdMultiply3(&epd1, &epd2, epd);
  }
}


/**Function********************************************************************

  Synopsis    [Computes arbitrary precision pow of base 2.]

  Description [Computes arbitrary precision pow of base 2.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdPow2Decimal(int n, EpDouble *epd)
{
  if (n <= EPD_MAX_BIN) {
    epd->type.value = pow((double)2.0, (double)n);
    epd->exponent = 0;
    EpdNormalizeDecimal(epd);
  } else {
    EpDouble    epd1, epd2;
    int         n1, n2;

    n1 = n / 2;
    n2 = n - n1;
    EpdPow2Decimal(n1, &epd1);
    EpdPow2Decimal(n2, &epd2);
    EpdMultiply3Decimal(&epd1, &epd2, epd);
  }
}


/**Function********************************************************************

  Synopsis    [Normalize an arbitrary precision double value.]

  Description [Normalize an arbitrary precision double value.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdNormalize(EpDouble *epd)
{
  int           exponent;

  if (IsNanOrInfDouble(epd->type.value)) {
    epd->exponent = 0;
    return;
  }

  exponent = EpdGetExponent(epd->type.value);
  if (exponent == EPD_MAX_BIN)
    return;
  exponent -= EPD_MAX_BIN;
  epd->type.bits.exponent = EPD_MAX_BIN;
  epd->exponent += exponent;
}


/**Function********************************************************************

  Synopsis    [Normalize an arbitrary precision double value.]

  Description [Normalize an arbitrary precision double value.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdNormalizeDecimal(EpDouble *epd)
{
  int           exponent;

  if (IsNanOrInfDouble(epd->type.value)) {
    epd->exponent = 0;
    return;
  }

  exponent = EpdGetExponentDecimal(epd->type.value);
  epd->type.value /= pow((double)10.0, (double)exponent);
  epd->exponent += exponent;
}


/**Function********************************************************************

  Synopsis    [Returns value and decimal exponent of EpDouble.]

  Description [Returns value and decimal exponent of EpDouble.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdGetValueAndDecimalExponent(EpDouble *epd, double *value, int *exponent)
{
  EpDouble      epd1, epd2;

  if (EpdIsNanOrInf(epd))
    return;

  if (EpdIsZero(epd)) {
    *value = 0.0;
    *exponent = 0;
    return;
  }

  epd1.type.value = epd->type.value;
  epd1.exponent = 0;
  EpdPow2Decimal(epd->exponent, &epd2);
  EpdMultiply2Decimal(&epd1, &epd2);

  *value = epd1.type.value;
  *exponent = epd1.exponent;
}

/**Function********************************************************************

  Synopsis    [Returns the exponent value of a double.]

  Description [Returns the exponent value of a double.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
EpdGetExponent(double value)
{
  int           exponent;
  EpDouble      epd;

  epd.type.value = value;
  exponent = epd.type.bits.exponent;
  return(exponent);
}


/**Function********************************************************************

  Synopsis    [Returns the decimal exponent value of a double.]

  Description [Returns the decimal exponent value of a double.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
EpdGetExponentDecimal(double value)
{
  char  *pos, str[24];
  int   exponent;

  sprintf(str, "%E", value);
  pos = strstr(str, "E");
  sscanf(pos, "E%d", &exponent);
  return(exponent);
}


/**Function********************************************************************

  Synopsis    [Makes EpDouble Inf.]

  Description [Makes EpDouble Inf.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdMakeInf(EpDouble *epd, int sign)
{
  epd->type.bits.mantissa1 = 0;
  epd->type.bits.mantissa0 = 0;
  epd->type.bits.exponent = EPD_EXP_INF;
  epd->type.bits.sign = sign;
  epd->exponent = 0;
}


/**Function********************************************************************

  Synopsis    [Makes EpDouble Zero.]

  Description [Makes EpDouble Zero.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdMakeZero(EpDouble *epd, int sign)
{
  epd->type.bits.mantissa1 = 0;
  epd->type.bits.mantissa0 = 0;
  epd->type.bits.exponent = 0;
  epd->type.bits.sign = sign;
  epd->exponent = 0;
}


/**Function********************************************************************

  Synopsis    [Makes EpDouble NaN.]

  Description [Makes EpDouble NaN.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdMakeNan(EpDouble *epd)
{
  epd->type.nan.mantissa1 = 0;
  epd->type.nan.mantissa0 = 0;
  epd->type.nan.quiet_bit = 1;
  epd->type.nan.exponent = EPD_EXP_INF;
  epd->type.nan.sign = 1;
  epd->exponent = 0;
}


/**Function********************************************************************

  Synopsis    [Copies a EpDouble struct.]

  Description [Copies a EpDouble struct.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
void
EpdCopy(EpDouble *from, EpDouble *to)
{
  to->type.value = from->type.value;
  to->exponent = from->exponent;
}


/**Function********************************************************************

  Synopsis    [Checks whether the value is Inf.]

  Description [Checks whether the value is Inf.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
EpdIsInf(EpDouble *epd)
{
  return(IsInfDouble(epd->type.value));
}


/**Function********************************************************************

  Synopsis    [Checks whether the value is Zero.]

  Description [Checks whether the value is Zero.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
EpdIsZero(EpDouble *epd)
{
  if (epd->type.value == 0.0)
    return(1);
  else
    return(0);
}


/**Function********************************************************************

  Synopsis    [Checks whether the value is NaN.]

  Description [Checks whether the value is NaN.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
EpdIsNan(EpDouble *epd)
{
  return(IsNanDouble(epd->type.value));
}


/**Function********************************************************************

  Synopsis    [Checks whether the value is NaN or Inf.]

  Description [Checks whether the value is NaN or Inf.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
EpdIsNanOrInf(EpDouble *epd)
{
  return(IsNanOrInfDouble(epd->type.value));
}


/**Function********************************************************************

  Synopsis    [Checks whether the value is Inf.]

  Description [Checks whether the value is Inf.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
IsInfDouble(double value)
{
  EpType val;

  val.value = value;
  if (val.bits.exponent == EPD_EXP_INF &&
      val.bits.mantissa0 == 0 &&
      val.bits.mantissa1 == 0) {
    if (val.bits.sign == 0)
      return(1);
    else
      return(-1);
  }
  return(0);
}


/**Function********************************************************************

  Synopsis    [Checks whether the value is NaN.]

  Description [Checks whether the value is NaN.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
IsNanDouble(double value)
{
  EpType        val;
  
  val.value = value;
  if (val.nan.exponent == EPD_EXP_INF &&
      val.nan.sign == 1 &&
      val.nan.quiet_bit == 1 &&
      val.nan.mantissa0 == 0 &&
      val.nan.mantissa1 == 0) {
    return(1);
  }
  return(0);
}


/**Function********************************************************************

  Synopsis    [Checks whether the value is NaN or Inf.]

  Description [Checks whether the value is NaN or Inf.]

  SideEffects []

  SeeAlso     []

******************************************************************************/
int
IsNanOrInfDouble(double value)
{
  EpType        val;

  val.value = value;
  if (val.nan.exponent == EPD_EXP_INF &&
      val.nan.mantissa0 == 0 &&
      val.nan.mantissa1 == 0 &&
      (val.nan.sign == 1 || val.nan.quiet_bit == 0)) {
    return(1);
  }
  return(0);
}

ABC_NAMESPACE_IMPL_END