Subject: New Math Library!
To: None <>
From: J.T. Conklin <>
List: current-users
Date: 02/10/1994 14:21:50
I've completed a first cut on a fdlibm derived math library.  Sun's
code works fine (even better than our old library on i386 machines
without math coprocessors); and I've converted the i387 & mc68881
support for our existing math lib.  

All that needs to be done is to integrate it into -current, and to
update our manpages to reflect the new library (Sun didn't provide
manual pages).

I have no idea what Sun's plans are with regards to future support of
fdlibm.  The Readme file seems to indicate that "float" versions of
the functions (sinf, cosf, etc.) will be done in a future release, but
it is very possible that Sun won't release anything more and we're on
our own.  If so, we're no worse off than we are today.

I'm convinced that integrating fdlibm is the correct thing to do, even
if it only supports IEEE math.  I don't see anyone jumping to do a VAX
or Tahoe port.  Unless someone complains really violently (Most of you
have known what I've been up to for a long while, so you've had your
chance), fdlibm is going to be integrated into NetBSD.

The only question that remains is how.  Since there's at least a
chance that Sun will release a new version of fdlibm, I'm thinking it
may be a good idea to keep the fdlibm source separate from the rest of
the library.

Something like this:

	     |	      |	         
	     |	      |-mc68881--
	     |	      |	       
	     |	      \-etc-- 

Any comments?