jmk@asr2.UUCP (06/07/87)
I goofed. But I did find out what having a math co-processor can do for you. When I repeated my orbit-prediction benchmark on a another 6300+, the results were much different (192 seconds vs. 22 reported previously). Sure enough, I've found that the original machine had an 80287 installed. Of course, it is not fair to compare the ST's performance to a machine with a co-processor. However, I think the ST should in theory be able to beat 192 seconds. Several people have already volunteered to try the program on various machines/compilers (including the Amiga with 68020), and I will post the results. The updated chart is: machine compiler run time (s) accuracy =========== ============ ======== ========== 520ST Lattice 3.03.01 480 very good 6300+ Microsoft 192 excellent 6300+/80287 Microsoft 22 excellent VAX 11/785 Unix V cc 4 excellent Amiga 1000 Manx -lm 25 poor Joe Knapp ihnp4!asr2!jmk AT&T Bell Labs Columbus OH (614)860-3547