home *** CD-ROM | disk | FTP | other *** search
- Path: cpca3.uea.ac.uk!news
- From: m.melton@uea.ac.uk (Mark Melton)
- Newsgroups: comp.sys.amiga.misc,comp.sys.amiga.hardware,comp.sys.amiga.advocacy
- Subject: Re: WAS: Motorola pricelist Now: Multiple CPUs
- Date: Sun, 21 Jan 1996 21:39:12 GMT
- Organization: University of East Anglia
- Distribution: world
- Message-ID: <4dubnr$ioi@cpca3.uea.ac.uk>
- References: <wfblanDKJ8Cw.1tn@netcom.com> <aTmosh.0qfe@amiga.ow.nl> <96017.131924IO92257@MAINE.MAINE.EDU> <4dor2g$2er@tbd120.tbd.ford.com> <4dtg98$6f1@news.ox.ac.uk>
- NNTP-Posting-Host: nc2g.nca.uea.ac.uk
- X-Newsreader: Forte Free Agent 1.0.82
-
- sjoh0358@sable.ox.ac.uk (Ian Parkinson) wrote:
-
- >With regard to having lots of processors, has anybody seen Acorn's RiscPC?
- >Some chaps from Acorn came and demonstrated it to us - they claim that it
- >is able to utilise multiple processors.
-
- >Even more impressively, they can use different types of chip - although at
- >the moment cards are only available for the ARM processor and for 486's.
-
- >This, I admit, made me consider (only for a second) dumping my Amiga... (Note:
- >I've always been impressed by Acorn, so this wasn't too unnatural!)
-
- >However, when I asked them how use of the multiple CPUs is implemented from a
- >software level, they said that it is essentially up to the individual tasks
- >to secure an extra processor and use it. I got the impression that multitasking
- >is performed just on one chip, but the programs can 'get' an extra chip and
- >use it exclusively. If this fails (i.e. because there aren't enough chips)
- >then they would have to fall back to normal multitasking.
-
- >This, IMO, is not good. Multitasking should be implemented on each CPU, i.e.
- >the tasks should be shared around the available processors. The idea is that
- >when any one CPU's current task 'runs out of time' it takes over any task (of
- >high enough priority) currently waiting.
-
- >I don't pretend to know anything about the hardware side of this sort of
- >implementation - it must be possible because Acorn have done it - but I don't
- >see any enormous difficulties on the OS side - if anybody would care to
- >contradict me I'd be happy to hear it!
-
- >For the record, the chappies from Acorn half-hinted to me that they were
- >planning this sort of system for RiscOS, but I got the impression that it
- >may have been sales pitch, and they didn't really know what they were talking
- >about...
-
- >If AT can produce something along these lines the world would rejoice!
- Well having spent a number of hours revising this very suject, I`ll be
- happy to contradict you.
- I have a friend who got a RISC PC a few weeks ago. What would appear
- to be happening is this. The RISC PC software run`s on the ARM and
- the PC software run`s on the 486. This is technically speaking a
- multi processor machine, but if we are prepared to call this multi
- processing then the amiga already is a multi processor machine. The
- copper in the amiga is a processor, it has its own instruction set and
- executes a list of instructions. It does this independantly from the
- 680x0 processor. Also if a bridgeboard is fitted to an amiga this is
- also a multi processor system. The RISC PC is nothing new in this
- sense, still you know what marketing people are like!!!
- ("We`ve just had this great Idea, its called multi media!" Microsoft)
- True multiprocessing, in the modern sense or what is correctly called
- a "centralised shared-memory architecture", consists of a number of
- identical processors, using a shared memory space. The real problem
- is what does the programmer see when they write software for the
- machine. Do they see individual processor which run totally separate
- tasks? or does the user see one virtual machine? The latter is
- obviously the best model, as further processors can be added without
- the need to re write programs to take advantage of multiple
- processors.
- Anyway I could go on all day but I must get some revision done!!
-
- Mark
- M.Melton@uea.ac.uk
-
-
-
-