rknight@gec-rl-hrc.co.uk (Roger Knight (B21)) (03/09/91)
In article <klamer.668261007@mi.eltn.utwente.nl> klamer@mi.eltn.utwente.nl (Klamer Schutte -- Universiteit Twente) writes: >PS And a question for the better-informed: When running on a R160 + ARM3 > under RISCiX a sample program without very much floating point did > perform only at 10% of a Sun Sparcstation 1+ (rated 15 MIPS). > The ARM3 should be better than 1.5 (sun) MIPS, isn't he? > Unix overhead is not the answer as the SS1+ did run SunOs 4.1 against > berkely 4.3 for the ARM. Where does the difference come from? Probably 2 reasons: (1) any floating point seems to have a disastrous effect on an ARM processor, and (2) the X-Windows display grabs a lot of memory bandwidth - it is like running in mode 21 on RISCos. Try switching to one of the VT220 emulation screens and it will go a lot faster (similarly RISCos goes like the wind if you switch the desktop to mode 0 ). We recently had a demo of the R260 and the benchmarks we ran on it put about equal to a SPARC IPC and faster than an APOLLO DN4500 on integer only (written in C of course!!). The benchmark was a quick hack at a souped up PCW integer test. Approx. timings were: R260 : 15.5 secs; SPARC IPC : 16.1 secs; Apollo DN4500 : 23.2 secs The R260 times went down by about 3 seconds by switching to a VT220 emulator display (all the machines were fairly idle, but still multi- user and running TCP/IP, NFS etc.). Putting a floating point divide into the loop didn't effect the SUN or Apollo but slowed the Acorn down by about 5 times. All I can do is repeat a quote from an article in comp.misc a few months back: "- Real programmers scorn floating point arith. The decimal point was invented for pansy bed-wetters who can't think big" -Rog. ( rknight@gec-rl-hrc.co.uk ) DISCLAIMER: I said it, not my company.