davidsen@sixhub.UUCP (Wm E. Davidsen Jr) (01/30/91)
In article <SIMON.91Jan24200351@lisp5.first.gmd.de> dieter@gmdtub.UUCP writes: | Are there any statistical data available on communication and/or memory | access patterns of parallel, non-numerical applications ? Since this reply is a good bit off the question, I'll post rather than mail. One interesting use of multiprocessing is for doing word processing. ie. soelim foo | tbl | eqn | troff | lpr -Pcat type stuff. Very low memory use, lots of i/o, but in general extremely CPU intensive and subject to having each process in its own processor. We do a lot of it at my work site, were running two VAXen and three Suns dedicated, plus Elan Eroff on about 100 PCs. We *like* troff! We have some large jobs (read books) which used to run 405 hours on a VAX 11-780. While you don't need a job that large as a benchmark, a few chapters give useful numbers. Large computers without byte addressing don't do all that well on troff, not surprisingly, and system which can put the pipe overhead into still another CPU do very well. I don't have figures to share, but perhaps it's something you might want to try on a highly parallel machine. I can probably come up with some large non-proprietary stuff for a "standard" test case. UNIX specific, of course. -- bill davidsen - davidsen@sixhub.uucp (uunet!crdgw1!sixhub!davidsen) sysop *IX BBS and Public Access UNIX moderator of comp.binaries.ibm.pc and 80386 mailing list "Stupidity, like virtue, is its own reward" -me