dan@rna.UUCP (Dan Ts'o) (05/29/90)
Does anyone have a C routine to convert from IEEE floating point to VAX D format floating point ? Failing that or anything similar, does anyone have a concise description of the IEEE floating point format so I can write my own routine... Please email responses. Thanks. Cheers, Dan Ts'o 212-570-7671 Dept. Neurobiology dan@rna.rockefeller.edu Rockefeller Univ. ...cmcl2!rna!dan 1230 York Ave. rna!dan@nyu.edu NY, NY 10021 tso@rockefeller.arpa tso@rockvax.bitnet
mf@uunet.UU.NET (Michel Fingerhut) (05/29/90)
There is a set of undocumented such routines in UWS 2.0 (see release notes p. 78).
tody@noao.edu (Doug Tody CCS) (06/03/90)
From article <1023@rna.UUCP>, by dan@rna.UUCP (Dan Ts'o): > > Does anyone have a C routine to convert from IEEE floating point > to VAX D format floating point ? This depends upon what cpu you want to perform the conversion on. If you want to convert to or from IEEE on a VAX cpu then I have some VAX assembler routines which can do this (more efficiently than in C, due to the need to bit shift a 64 bit quantity). > Failing that or anything similar, does anyone have a concise > description of the IEEE floating point format so I can write my own routine... The main reference is the IEEE floating standard document. I don't have the document number handy, but you can order it from IEEE. An IEEE double is 64 bits like a VAX D double, but the exponent is 11 bits rather than 8 for VAX D format (VAX G format is also 11 bits), there is a *4 (or /4) factor due to a difference in the normalization of the mantissa, and the bytes are spped. Converting IEEE single to VAX F is much simpler, all you do is byte swap and multiply or divide by 4.0. -- Doug Tody, National Optical Astronomy Observatories, Tucson AZ, 602-325-9217 UUCP: {arizona,decvax,ncar}!noao!tody or uunet!noao.edu!tody Internet: tody@noao.edu SPAN/HEPNET: NOAO::TODY (NOAO=5355)
bright@Data-IO.COM (Walter Bright) (06/05/90)
From article <1023@rna.UUCP>, by dan@rna.UUCP (Dan Ts'o):
< Does anyone have a C routine to convert from IEEE floating point
< to VAX D format floating point ?
Here's one I wrote to convert from VAX to PC 64-bit doubles. You could
easily reverse it to go the other way.
main()
{
double d1,d2;
long *p;
while (1)
{ scanf(" %lf",&d1);
p = (long *) &d1;
printf("x%08lx,x%08lx\n",p[1],p[0]);
vaxdbl_to_pcdbl(&d1,&d2);
p = (long *) &d2;
printf("x%08lx,x%08lx\n",p[1],p[0]);
}
}
vaxdbl_to_pcdbl(pvax,ppc)
unsigned long *pvax,*ppc;
{
unsigned sign,exponent;
unsigned long fraction;
sign = (*pvax & 0x8000) != 0;
exponent = (*pvax >> 7) & 0xFF;
fraction = ((*pvax & 0x7F) << 16) | (*pvax >> 16);
printf("sign = %d, exponent = x%x, fraction = x%lx\n",
sign,exponent,fraction);
if (sign == 0 && exponent == 0)
{ ppc[0] = 0;
ppc[1] = 0;
return;
}
/* Put sign in result */
ppc[1] = (unsigned long) sign << 31;
/* Compute new exponent */
exponent += 0x3FF - 0x80 - 1;
ppc[1] |= (unsigned long) exponent << 20;
/* OR in fraction */
ppc[1] |= (fraction >> 3);
ppc[0] = (fraction << 29) | (pvax[1] >> 3);
}wallyk@tekfdi.FDI.TEK.COM (Wally Kramer) (06/05/90)
This seems to be a subject which crops up from time to time. I have four routines which perform the following conversions: DEC F_FLOAT to/from IEEE ShortReal DEC D_FLOAT to/from IEEE LongReal They are tested using the VAXC C compiler (from DEC), Turbo C 2.0, and Green Hills C compiler for the 386. As written, they do not appear to work on Sun compilers for the 68000, but should be easy to modify for this purpose. They are written in ANSI C without the use of floating point instructions (that is, no float or double declarations--just structure manipulations). There is sufficient documentation to write the remaining conversions, or to learn just about anything you want about either DEC or IEEE formats. I'll be happy to email a copy to anyone who asks. (I prefer Email requests. Use the phone only if your mailer can't reach me.) wallyk@tekfdi.fdi.tek.com (Wally Kramer) 503 627 2363 Contractor from Step Technology, Inc. 503 244 1239
craigs@cognos.UUCP (Craig Statchuk) (06/07/90)
>From article <1023@rna.UUCP>, by dan@rna.UUCP (Dan Ts'o): >> >> Does anyone have a C routine to convert from IEEE floating point >> to VAX D format floating point ? > I had to do this a while back. I found it easiest to convert IEEE -> G_FLOAT ->D_FLOAT. As mentioned in another response all you have to do swap bytes and change the exponent to go from IEEE to G_FLOAT. The VAX RTL has the necessary routine to change from G_FLOAT to D_FLOAT (and has another one if you wish to go back). Here's some code that will do the translation on the (using bit fields yeah!): union G_float_template { double asDouble; struct { unsigned long b1:16,b2:16,b3:16,b4:16; } asBits; struct { unsigned long mantissaL:20,exponent:11,sign:1,mantissaH:32; } asFields; }; union G_float_template value; unsigned short s_temp; double result; extern double MTH$CVT_G_D(); ... /* put some value into the template -- assume it is really in IEEE format */ value.asDouble = (IEEE) 1234.5 /* exchange upper 16 bits with lower 16 bits in each word */ s_temp = value.asBits.b1; value.asBits.b1 = value.asBits.b2; value.asBits.b2 = s_temp; s_temp = value.asBits.b3; value.asBits.b3 = value.asBit.b4; value.asBits.b4 = s_temp; /* normalize exponent bias */ if (value.asDouble != 0.0) net_value.asFields.exponent -= 2; /* convert to D_FLOAT */ result = MTH$CVT_G_D(&value.asDouble); Easy eh? /CS -- Craig Statchuk USENET : uunet!mitel!sce!cognos!craigs Cognos Incorporated INTERNET : craigs%cognos.uucp@uunet.uu.net 3755 Riverside Dr. MaBellNET: (613) 738-1440 Ottawa, Ontario K1G 3Z4 FaxNET : (613) 738-0002