[comp.sys.next] Speed of 040 with math functions

erik@zeus.uucp (Erik Schumacher) (02/23/91)

I saw some postings regarding the speed of the 040 regarding math functions.
However, I did not understand them very well and would like to ask for some
more explanation.  My expirience was, that my Next station was much slower
than I had expected from the MIPS and FLOPS numbers.  I compared the times
on my machine with times on a IBM RISC 320 and it turned out that that the
RISC was between 3 and 14 times faster, depending on the program.  I also
noted that my system times took a substantial portion of the overall time
(up to 1/2), whereas the system time on the RISC was always close to zero.
From the MIPS and FLOPS numbers, the RISC should ony be 3 to 4 times faster.
Any explanations and hints to speed up the 040 will be very helpful.  
Thank you.


Erik Schumacher
Optical Sciences Center
University of Arizona

erik@zeus.opt-sci.arizona.edu