[comp.sys.mac.misc] SGIL "Fast Math" - Just Say No?

barvian@ece.cmu.edu (Scott Barvian) (05/03/91)

Dammit, I *knew* it was too good to be true.  I've gotten messages that
the "Miracle Math Accelerator from Hell, er.. France" causes problems in
the following applications:
- HyperCard
- Managing Your Money (isn't that conveeenient?)
- DeltaGraph (my own personal finding- graphs come out unreadable, and the
              spreadsheet gives results that are completely out to lunch)

Remember when your grandfather said "there's no free lunch"?  Maybe the
old guy was right after all.  I'm putting SGIL in the init graveyard for now.

Sorry to get everyone interested in this for nothing.  I'm sending this 
file on to the umich archive guys to get this info stored with the file
(unless they decide to pull it).

--
==== Scott Barvian ============ Department of Electrical/Computer Eng. ===
==== barvian@ece.cmu.edu ====== Carnegie Mellon Univ., Pittsburgh, PA  ===

n67786@cc.tut.fi (Tero Nieminen) (05/04/91)

In article <BARVIAN.91May2185034@dart.ece.cmu.edu> barvian@ece.cmu.edu (Scott Barvian) writes:

   Dammit, I *knew* it was too good to be true.  I've gotten messages that
   the "Miracle Math Accelerator from Hell, er.. France" causes problems in
   the following applications:
   - HyperCard
   - Managing Your Money (isn't that conveeenient?)
   - DeltaGraph (my own personal finding- graphs come out unreadable, and the
		 spreadsheet gives results that are completely out to lunch)

   Remember when your grandfather said "there's no free lunch"?  Maybe the
   old guy was right after all.  I'm putting SGIL in the init graveyard for now.

   Sorry to get everyone interested in this for nothing.  I'm sending this 
   file on to the umich archive guys to get this info stored with the file
   (unless they decide to pull it).

Also worth remembering is that the other fast math inits exhibit same
kind of behaviour as the ones discussed here lately. I ran into this
with Radius FastMath init and ClarisCAD, where it caused abnormal
rounding errors and incorrect mesurement. It was clearly visible even at
drawing time when the picture was magnified enough. So beware..  :)
-- 
   Tero Nieminen                    Tampere University of Technology
   n67786@cc.tut.fi                 Tampere, Finland, Europe