lawsonse@vttcf.cc.vt.edu (Shannon Lawson) (04/07/91)
I figured out the other problem I was having. It turns out that Aztec C
doesn't like defines to be spread to extra lines using "\" as a continuation
character.
Here's another oddity:
My A3000UX has the standard 68030/68882 configuration. I am using the following
command lines for cc/ln:
cc -c2 -f8 foo.c
^ ^
| |
| Generate 68881 compatible code (should work for an '882, right?)
Generate 68020 compatible code (again OK for '030?)
ln foo.o -lm8 -lc
^ ^
| |
| Standard C library
Motorola 68881 floating point library
Everything compiles/assembles/links just fine. My question is this:
How come the math coprocessor version of foo runs about 2.5 times SLOWER
than if I use the Motorola Fast Floating Point emulation options?
(-ff for "cc", and -lmf for "ln")
Motorola FFP only uses 32-bit math, so I guess it's conceivable that the
extra precision used by the math coprocessor would introduce some overhead.
Somehow, I just figured the coprocessor would be the way to go.
Am I missing something here?
Makes me wish I had Turbo C for the Amiga! Sigh...
Borland, are you listening? :-/
--
Shannon Lawson || Amiga Student on Campus Consultant
Senior, Computer Engineering || Virginia Polytechnic Institute
Harry Lynde Bradley Department of || & State University
Electrical Engineering || E-Mail: lawsons@csgrad.cs.vt.edu
Disclaimer: If I put my foot in my mouth, I can hardly hold someone else
responsible, now can I?