rs4u#@ANDREW.CMU.EDU (Richard Siegel) (05/07/87)
The MS Fortran floating-point routines handle single and double precision
real numbers -- 32 and 64 bits, in the same IEEE standard
format as SANE does. The Mac's SANE routines coerce everything to 80
bits when operating; this is why Extended variables run faster
in computations that REAL variables or Double variables -- up to
a factor of 4 for simple math.
If you use that Pascal REAL type now, you won't notice any loss of precision
if you use the Fortran math routines.
--Rich
Richard M. Siegel
Materials Characterization Instrumentation Section
Mail Stop 231
NASA/Langley Research Center
Hampton, Virginia 23665
(804) 865-3036
Arpanet: rs4u@andrew.cmu.edu
Uucp: {your fave gateway}!seismo!andrew.cmu.edu!rs4u
Disclaimer? I don't even KNOW 'er!wmcb@ecsvax.UUCP (William C. Bauldry) (05/07/87)
Thanks for the info. Since I generally use extended types, it doesn't look like I'd want to switch. The reason for using extended is that the recurrences I use are extremely unstable non-linear and blow up quickly - much too much for single - using extended only helps a little but... Thanks again. Bill