I've just reviewed AN575.  Like the other math
routines they very cavilerly used registers
declaring them with equates only.  You can
corrupt them VERY easily and never know it.
Been there and done that.  Take a close stare
at the include files and make sure your variables
don't disturb (overwrite) their variables.

I've successfully used some of the other math
routines and found the above variable cautions
to be the only caveats.

GL,
Walt

-----Original Message-----
From: Janus Ternullo [mailto:Jman1@WORLD.STD.COM]
Sent: Thursday, December 02, 1999 5:02 PM
To: PICLIST@MITVMA.MIT.EDU
Subject: floating point routine errors


Has anyone used the 32-bit floating point routines in AN575 with success
(in particular FPM32 and FPD32)?  After various trials, I found that the
code seems to confuse the exponent variables EXP and AEXP for the AARGB
exponent, when the instructions say to use AEXP.  I found what appears
to be a workaround for FPM32 (assign the exponent to both AEXP and EXP),
but haven't had the same luck with FPD32.  I may be using the routines
wrong, but I don't think so.  Any solutions or insight would be
appreciated (a source for corrected code, alternate code, etc).  I can
use solutions in either assembly or C (MPLAB-C17).  Ultimately I would
like to get the square root routine in AN660 working.

Thanks and Regards,
Janus