mirror of https://github.com/tongzx/nt5src
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
317 lines
9.4 KiB
317 lines
9.4 KiB
/*-----------------------------------------------------------------------------
|
|
*
|
|
* cvtibmf.c : IBM float/double <-> IEEE float/double conversion
|
|
*
|
|
*+++
|
|
*
|
|
* Copyright (c) Software AG 1996,1998. All rights reserved.
|
|
*
|
|
*---
|
|
*
|
|
* License:
|
|
*
|
|
* "According to the DCOM Porting Agreement Software AG grants to Microsoft
|
|
* an irrevocable, unlimited, royalty free license to use and market the
|
|
* enclosed piece of software code in source and object format for
|
|
* the purposes of Microsoft. 17-April-1998."
|
|
*
|
|
*----------------------------------------------------------------------------*/
|
|
|
|
#include <float.h>
|
|
|
|
#include <rpc.h>
|
|
#include <rpcndr.h>
|
|
#include "winerror.h"
|
|
|
|
/*
|
|
* Convert floating point numbers from IBM/370 to IEEE representation or vice versa.
|
|
*
|
|
* Synopsis:
|
|
*
|
|
* void cvt_ibm_f_to_ieee_single(ULONG *ulFP);
|
|
* void cvt_ibm_d_to_ieee_double(ULONG *ulFP);
|
|
* void cvt_ieee_single_to_ibm_f(ULONG *ulFP);
|
|
* void cvt_ieee_double_to_ibm_d(ULONG *ulFP);
|
|
*
|
|
* Note:
|
|
*
|
|
* Overflow/Underflow during conversion results in RpcRaiseException(RPC_S_FP_OVERFLOW/
|
|
* RPC_S_FP_UNDERFLOW).
|
|
*
|
|
*/
|
|
|
|
/* *******************************************************************************
|
|
*
|
|
* Floating point representations:
|
|
*
|
|
* ------------------------
|
|
* IBM/370 single precision
|
|
* ------------------------
|
|
*
|
|
* xxxx.xxxx xxxx.xxxx xxxx.xxxx xxxx.xxxx
|
|
* s|-exp--| |--------fraction-----------|
|
|
* (7) (24)
|
|
*
|
|
* value = (-1)**s * 16**(e - 64) * .f range = 5.4E-79 ... 7.2E+75
|
|
*
|
|
* *******************************************************************************
|
|
*
|
|
* ---------------------
|
|
* IEEE single precision
|
|
* ---------------------
|
|
*
|
|
* xxxx.xxxx xxxx.xxxx xxxx.xxxx xxxx.xxxx
|
|
* s|--exp---||-------fraction-----------|
|
|
* (8) (23)
|
|
*
|
|
* value = (-1)**s * 2**(e - 127) * 1.f range = 1.2E-38 ... 3.4E+38
|
|
*
|
|
* *******************************************************************************
|
|
*
|
|
* ------------------------
|
|
* IBM/370 double precision
|
|
* ------------------------
|
|
*
|
|
* xxxx.xxxx xxxx.xxxx xxxx.xxxx xxxx.xxxx yyyy.yyyy yyyy.yyyy yyyy.yyyy yyyy.yyyy
|
|
* s|-exp--| |-------------------------------fraction----------------------------|
|
|
* (7) (56)
|
|
*
|
|
* value = (-1)**s * 16**(e - 64) * .f range = 5.4E-79 ... 7.2E+75
|
|
*
|
|
* *******************************************************************************
|
|
*
|
|
* ---------------------
|
|
* IEEE double precision
|
|
* ---------------------
|
|
*
|
|
* xxxx.xxxx xxxx.xxxx xxxx.xxxx xxxx.xxxx yyyy.yyyy yyyy.yyyy yyyy.yyyy yyyy.yyyy
|
|
* s|--exponent-| |-------------------------fraction-----------------------------|
|
|
* (11) (52)
|
|
*
|
|
* value = (-1)**s * 2**(e - 1023) * 1.f range = 2.2E-308 ... 1.8+308
|
|
*
|
|
* *******************************************************************************/
|
|
|
|
#if 1 /* We assume little endian for NT, this does not work: NDR_LOCAL_ENDIAN == NDR_LITTLE_ENDIAN */
|
|
# define HI 1 /* index of high order LONG */
|
|
# define LO 0 /* index of low order LONG */
|
|
#else
|
|
# define HI 0 /* index of high order LONG */
|
|
# define LO 1 /* index of low order LONG */
|
|
#endif
|
|
|
|
static float floatMin = FLT_MIN;
|
|
static float floatMax = FLT_MAX;
|
|
|
|
#define SIGN(src) (src[HI] & 0x80000000)
|
|
|
|
/* Convert IBM/370 "float" to IEEE "single" */
|
|
void cvt_ibm_f_to_ieee_single ( ULONG *ulFP )
|
|
{
|
|
ULONG ulFraction ;
|
|
LONG lExponent ;
|
|
|
|
/* in this special case we just keep the sign */
|
|
if ( ( *ulFP & 0x7fffffff ) == 0 )
|
|
{
|
|
return ;
|
|
}
|
|
|
|
/* fetch the exponent (excess-64 notation) and fraction */
|
|
lExponent = ( (*ulFP & 0x7f000000) >> 24) - 64 ;
|
|
ulFraction = *ulFP & 0x00ffffff ;
|
|
|
|
/* convert from "16**exponent" to "2**exponent" */
|
|
if ( lExponent >= 0 ) lExponent <<= 2 ;
|
|
else lExponent = -((-lExponent) << 2) ;
|
|
|
|
/* convert exponent for 24 bit fraction to 23 bit fraction */
|
|
lExponent -= 1;
|
|
|
|
/* normalize fraction */
|
|
if ( ulFraction )
|
|
{
|
|
while ( (ulFraction & 0x00800000) == 0 )
|
|
{
|
|
ulFraction <<= 1 ;
|
|
lExponent -= 1 ;
|
|
}
|
|
}
|
|
|
|
/* remove the implied '1' preceeding the binary point */
|
|
ulFraction &= 0x007fffff ;
|
|
|
|
/* convert exponent to excess-127 notation and store the number if the exponent is not out of range */
|
|
if ( (lExponent += 127) >= 255 )
|
|
*ulFP = SIGN(ulFP) | *((ULONG *)&floatMax) ; /* floating overflow */
|
|
else if ( lExponent <= 0 )
|
|
*ulFP = SIGN(ulFP) | *((ULONG *)&floatMin) ; /* floating underflow */
|
|
else
|
|
*ulFP = SIGN(ulFP) | (lExponent << 23) | ulFraction ;
|
|
}
|
|
|
|
/* Convert IBM/370 "double" to IEEE "double" */
|
|
void cvt_ibm_d_to_ieee_double ( ULONG* ulFP )
|
|
{
|
|
ULONG ulFraction[2] ;
|
|
LONG lExponent ;
|
|
|
|
/* in this special case we just keep the sign */
|
|
if ( (ulFP[HI] & 0x7fffffff) == 0 )
|
|
{
|
|
return ;
|
|
}
|
|
|
|
/* fetch the exponent (removing excess 64) and fraction */
|
|
lExponent = ( (ulFP[HI] & 0x7f000000) >> 24 ) - 64 ;
|
|
ulFraction[HI] = ulFP[HI] & 0x00ffffff ;
|
|
ulFraction[LO] = ulFP[LO] ;
|
|
|
|
/* convert from "16**exponent" to "2**exponent" */
|
|
if ( lExponent >= 0 ) lExponent <<= 2 ;
|
|
else lExponent = -((-lExponent) << 2);
|
|
|
|
/* normalize the fraction (to 57 bits) */
|
|
if ( ulFraction[HI] )
|
|
{
|
|
while ((ulFraction[HI] & 0x01000000) == 0)
|
|
{
|
|
ulFraction[HI] = ( ulFraction[HI] << 1 ) | ( ulFraction[LO] >> 31 ) ;
|
|
ulFraction[LO] = ulFraction[LO] << 1 ;
|
|
lExponent -= 1 ;
|
|
}
|
|
}
|
|
|
|
/* convert 57 bit fraction to 53 bit fraction and remove the implied '1' preceeding the binary point */
|
|
ulFraction[LO] = ( ulFraction[LO] >> 4 ) | ( ulFraction[HI] << 28 ) ;
|
|
ulFraction[HI] = ( ulFraction[HI] >> 4 ) & 0x000fffff ;
|
|
|
|
/* convert exponent to excess-1023 notation and store the number if the exponent is not out of range */
|
|
if ( (lExponent += 1023) >= 2047 )
|
|
RpcRaiseException ( RPC_S_FP_OVERFLOW ) ; /* should never happen */
|
|
else if ( lExponent <= 0 )
|
|
RpcRaiseException ( RPC_S_FP_UNDERFLOW ) ; /* should never happen */
|
|
else
|
|
{
|
|
ulFP[HI] = SIGN(ulFP) | (lExponent << 20) | ulFraction[HI] ;
|
|
ulFP[LO] = ulFraction[LO] ;
|
|
}
|
|
}
|
|
|
|
/* The following is not used in Windows NT */
|
|
|
|
#if 0
|
|
|
|
/* Convert IEEE "single" to IBM/370 "float" */
|
|
void cvt_ieee_single_to_ibm_f ( ULONG* ulFP )
|
|
{
|
|
ULONG ulFraction ;
|
|
LONG lExponent ;
|
|
|
|
/* in this special case we just keep the sign */
|
|
if ( (*ulFP & 0x7fffffff) == 0 )
|
|
{
|
|
return ;
|
|
}
|
|
|
|
lExponent = ((*ulFP & 0x7f800000) >> 23) - 127 ;
|
|
ulFraction = *ulFP & 0x007fffff ;
|
|
|
|
/* convert 23 bit fraction to 24 bit fraction */
|
|
ulFraction <<= 1 ;
|
|
|
|
/* restore the implied '1' which preceeded the IEEE binary point */
|
|
ulFraction |= 0x01000000 ;
|
|
|
|
/* convert from "2**exponent" to "16**exponent" (fraction is not normalized) */
|
|
if ( lExponent >= 0 )
|
|
{
|
|
ulFraction <<= (lExponent & 3) ;
|
|
lExponent >>= 2 ;
|
|
}
|
|
else
|
|
{
|
|
ulFraction >>= ((-lExponent) & 3) ;
|
|
lExponent = -((-lExponent) >> 2) ;
|
|
}
|
|
|
|
/* reduce fraction to 24 bits or less */
|
|
if ( ulFraction & 0x0f000000 )
|
|
{
|
|
ulFraction >>= 4 ;
|
|
lExponent += 1 ;
|
|
}
|
|
|
|
/* convert exponent to excess-64 notation and store the number if the exponent is not out of range */
|
|
if ( (lExponent += 64) > 127 )
|
|
RpcRaiseException ( RPC_S_FP_OVERFLOW ) ; /* should never happen */
|
|
else if ( lExponent < 0 )
|
|
RpcRaiseException ( RPC_S_FP_UNDERFLOW ) ; /* should never happen */
|
|
else
|
|
*ulFP = SIGN(ulFP) | (lExponent << 24) | ulFraction ;
|
|
}
|
|
|
|
/* Convert IEEE "double" to IBM/370 "double" */
|
|
void cvt_ieee_double_to_ibm_d ( ULONG* ulFP )
|
|
{
|
|
ULONG ulFraction[2] ;
|
|
LONG lExponent ;
|
|
LONG shift ;
|
|
|
|
/* in this special case we just keep the sign and the low word */
|
|
if ( (ulFP[HI] & 0x7fffffff) == 0 )
|
|
{
|
|
return ;
|
|
}
|
|
|
|
/* fetch the exponent (excess-1023 notation) and fraction */
|
|
lExponent = ((ulFP[HI] & 0x7ff00000) >> 20) - 1023 ;
|
|
ulFraction[HI] = ulFP[HI] & 0x000fffff ;
|
|
ulFraction[LO] = ulFP[LO] ;
|
|
|
|
/* convert 52 bit fraction to 56 bit fraction and restore the '1' which preceeds the IEEE binary point*/
|
|
ulFraction[HI] = ( ulFraction[HI] << 4 ) | ( ulFraction[LO] >> 28 ) | 0x01000000 ;
|
|
ulFraction[LO] = ulFraction[LO] << 4 ;
|
|
|
|
/* convert from "2**exponent" to "16**exponent" (fraction is not normalized) */
|
|
if ( lExponent >= 0 )
|
|
{
|
|
shift = lExponent & 3 ;
|
|
ulFraction[HI] = ( ulFraction[HI] << shift ) | ( ulFraction[LO] >> (32 - shift) ) ;
|
|
ulFraction[LO] = ulFraction[LO] << shift ;
|
|
lExponent >>= 2 ;
|
|
}
|
|
else
|
|
{
|
|
shift = (-lExponent) & 3 ;
|
|
ulFraction[LO] = ( ulFraction[LO] >> shift ) | ( ulFraction[HI] << (32 - shift) ) ;
|
|
ulFraction[HI] = ( ulFraction[HI] >> shift ) ;
|
|
lExponent = -((-lExponent) >> 2) ;
|
|
}
|
|
|
|
/* reduce fraction to 56 bits or less */
|
|
if ( ulFraction[HI] & 0x0f000000 )
|
|
{
|
|
ulFraction[LO] = ( ulFraction[LO] >> 4 ) | ( ulFraction[HI] << 28 ) ;
|
|
ulFraction[HI] = ( ulFraction[HI] >> 4 ) ;
|
|
lExponent += 1 ;
|
|
}
|
|
|
|
/* convert exponent to excess-64 notation and store the number */
|
|
if ( (lExponent += 64) > 127 )
|
|
{ /* we store the highest IBM float but we keep the sign ! */
|
|
ulFP[HI] = SIGN(ulFP) | 0x7FFFFFFF ;
|
|
ulFP[LO] = 0xFFFFFFFF ;
|
|
}
|
|
else if ( lExponent < 0 )
|
|
{ /* we store 0 but we keep the sign ! */
|
|
ulFP[HI] = SIGN(ulFP) ;
|
|
ulFP[LO] = 0 ;
|
|
}
|
|
else
|
|
{
|
|
ulFP[HI] = SIGN(ulFP) | (lExponent << 24) | ulFraction[HI] ;
|
|
ulFP[LO] = ulFraction[LO] ;
|
|
}
|
|
}
|
|
#endif
|