home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!uknet!gdt!aber!aberfa!pcg
- From: pcg@aber.ac.uk (Piercarlo Grandi)
- Newsgroups: comp.arch
- Subject: Re: No progress in parallelism? (was: History)
- Message-ID: <PCG.92Aug27124718@aberdb.aber.ac.uk>
- Date: 27 Aug 92 12:47:18 GMT
- References: <PCG.92Aug26005248@aberdb.aber.ac.uk> <1992Aug26.153025.1147@sctc.com>
- Sender: news@aber.ac.uk (USENET news service)
- Reply-To: pcg@aber.ac.uk (Piercarlo Grandi)
- Organization: Prifysgol Cymru, Aberystwyth
- Lines: 47
- In-Reply-To: smith@sctc.com's message of 26 Aug 92 15: 30:25 GMT
- Nntp-Posting-Host: aberdb
-
- On 26 Aug 92 15:30:25 GMT, smith@sctc.com (Rick Smith) said:
-
- smith> I think the future of computer architectures is in various forms
- smith> of parallelism not addressed by the von Neumann tradition.
-
- smith> And pcg@aber.ac.uk (Piercarlo Grandi) replied:
-
- pcg> Pious hope, I am afraid. Research in this field has been going on
- pcg> forever, and without much progress. Architectures here seem wedded
- pcg> to particular application shapes, and almost all variations have
- pcg> been tried out, and only a few have been found cost effective. But
- pcg> some particular application shapes are so important...
-
- smith> It should be clear from von Neumann's writings that
- smith> multiprocessing was never part of his model. However, I think it
- smith> is fair to say that the cost effectiveness of multiprocessor
- smith> configurations has been well established.
-
- Since at least the early sixties (and there were sort-of multiprocessors
- in the fifties), in some important application areas. And closely
- coupled multiprocessors are in the Von Neumann tradition, even if the
- man himself did not write about them. After all multiprocessing is not
- an architectural issue, it is an implementation issue. The architecture
- of a Von Neumann style multiprocessor is not in any way different from
- that of a Von Neumann uniprocessor.
-
- The interesting aspects of parallelism rather lie in large scale
- parallelism, which is something that requires architectures not in the
- Von Neumann tradition, and as to that the cost effectiveness has been
- domonstrated only in fairly narrow, if important, application domains.
-
- Such fairly narrow domains however are going to become fairly common, as
- multimedia and virtual reality percolate to the masses.
-
- smith> I think it's also fair to say that many of the "big issues" these
- smith> days (cache management comes immediately to mind) are driven by
- smith> parallelism problems.
-
- Again, nothing new. Many such issues have existed forever, even before
- caches. I am thinking of the GE 600 series system controllers, for
- example. Yes, there has been some progress, more interesting algorithms.
- Hardly anything new under the sun for computer architecture, and even
- for computer engineering.
- --
- Piercarlo Grandi | JNET: pcg@uk.ac.aber
- Dept of CS, University of Wales | UUCP: ...!mcsun!ukc!aber-cs!pcg
- Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@aber.ac.uk
-