home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!sun4nl!cwi.nl!dik
- From: dik@cwi.nl (Dik T. Winter)
- Newsgroups: comp.arch
- Subject: Re: Multiprocessors
- Message-ID: <8320@charon.cwi.nl>
- Date: 12 Dec 92 22:30:30 GMT
- References: <wilson.723024694@moonshine> <41747@sdcc12.ucsd.edu> <Bz5s9F.J32.2@cs.cmu.edu>
- Sender: news@cwi.nl
- Organization: CWI, Amsterdam
- Lines: 20
-
- In article <Bz5s9F.J32.2@cs.cmu.edu> lindsay+@cs.cmu.edu (Donald Lindsay) writes:
- > Oddly, although Crays have been multiprocessors for years, a lot of
- > programs only use one Cray processor. This is often largely a
- > function of incentive. One installation offered a big budget break to
- > parallel applications, and sure enough, almost everything suddenly
- > turned out to be parallelizable.
-
- I do not think it is very odd. If there is no incentive by budget breaks
- or things like that, a program which has been parallelized will cost you
- more than a program not parallelized. The program takes more raw CPU time,
- although it is done faster.
-
- This is always a problem in places where budgets are used. I know of
- programs for the Cyber 205 that vectorized extremely well but should be
- run scalar. The reason was that the vectorized version required more
- memory and you had to pay for that, so much more that it completely
- eliminated the advantage of the faster result in CPU time.
- --
- dik t. winter, cwi, kruislaan 413, 1098 sj amsterdam, nederland
- home: bovenover 215, 1025 jn amsterdam, nederland; e-mail: dik@cwi.nl
-