rey@psuecl.bitnet (11/16/89)
I have a question for anyone with experience with the 68882 math coprocessor.
The company that I work for is interested in producing a real time navigation
emulator on a desktop computer. Computation power is a seriou n
as we will be running the emulation and a 12 state Kalman filter in real
time (as well as display programs and interrupt driven software to interface
with external instruments).
Initially, the folks here want to use a 386/387 based PC. I am just begining
to check out the 80387 vs the 68882 FPU. The 882 appears to support
concurrent math operations and I'm not sure if the 387 can do the same.
The multiplication times for the 882 is a bit longer than the 387 (95 v
70 clocks) but if the '882 can run two concurrent (head to tail) operations
can one effectively halve the 95 clock counts?
My questions are:
1) Is one FPU appreciably faster than in real world applications?
2) Do any of the available C compilers for the Amiga take full advantage
of the '882s ability to run concurrent floating point operations.
3) What kind of performance benchmarks have been recorded on math intensive
applications with various C compilers.
On a completely different subject has anyone used the combination of the
Amiga side (supplimented with an 030) and the 286 bridgecard. I will need
the AT slots to provide an interface to interface cards which talk to
standard aircraft instrument busses (i.e. 1553 and ARINC). How can one
pass messages from the bridgecard to the Amiga side?
If you can help me out, maybe we can get Amy into a commercial product (or
at least we can try!). Please e-mail your insights to:
REY@PSUECL.BITNET
-or-
REY@PSU.ECL.EDU
Thanks to all in advance
Bobhill@evax.arl.utexas.edu (Anthony Adam Hill) (11/20/89)
Bob,
Check back issues of MIPS I beleive they had a math co-processor "shoot out"
a while back. If the 1553 and ARINC are IEEE compliant busses there is a product that will allow you to communicate with them.
Thats all I know.....
adam hill