Van Snyder wrote:
>
> Concerning ERFC from Numerical Recipes, Alvaro Fernandez wrote:
>
> > The error appears symmetrically about 0 in certain parts of the
> > domain x=[-5 5] only (around +/- 1.22 to +/-2.3. I'll hunt for a bug
> > of my own.
>
> Why hunt for a bug when it's so much easier to get reliable software,
> free either from www.netlib.org or gams.nist.gov. If you want support,
> too, contact nag.co.uk or http://www.vni.com/products/imsl/
>
I suppose stuff on netlib is more relaible, but as easier to use...
I have been fighting with various routines from netlib, all of of
which either compile but don't run properly because they're missing
routines (even downloading dependencies) or don't compile at all,
due to obscure or obsolete syntax. For exampel, I'm trying to use
ERFC from
http://gams.nist.gov/serve.cgi/Module/CMLIB/ERFC/187/
and picked fullsource. Try as I might, I can't seem to pick machine
constants that it likes. I'm running on a SUN Ultra, and F90
doesn't like any of the constants given for the SUN by the R1MACH
routine in ERFC.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|