home *** CD-ROM | disk | FTP | other *** search
- Path: nh.acorn.co.uk!uknet!str-ccsun!not-for-mail
- From: nbc@vulture.dmem.strath.ac.uk (Neil Brendan Clark)
- Newsgroups: comp.sys.amiga.misc
- Subject: Re: Acorn RiscPC --- a thought?
- Date: 19 Feb 1996 19:33:29 -0000
- Organization: University of Strathclyde
- Message-ID: <4gaja9$i1o@vulture.dmem.strath.ac.uk>
- References: <4fdglm$b8s@vulture.dmem.strath.ac.uk> <1996Feb12.152844.28437@leeds.ac.uk> <4fo381$c0t@vulture.dmem.strath.ac.uk> <1996Feb14.125120.13466@leeds.ac.uk>
- NNTP-Posting-Host: vulture.dmem.strath.ac.uk
-
- J M Oldak <csxjmo@scs.leeds.ac.uk> wrote:
- >Neil Brendan Clark writes:
- >>J M Oldak <csxjmo@scs.leeds.ac.uk> wrote:
- >>>Neil Brendan Clark writes:
- >>
- >But it also means that, at the end of the day, the applications are more
- >efficient.
-
- No, it does not. You seem to have the idea that a pre-emptive MT system
- incurs some sort of vast overhead. This is simply not the case - the overhead
- is insignificant. It also means that applications are less portable. I have
- already gone into reasons why cooperative MT is *less* efficient - in a sense
- it could even be said to "poll".
-
- >It might take a bit longer in development,
-
- i.e. less real devlopment gets done. One of the reasons for things like
- pre-emptive MT (PMT?), virtual memory, and standard libraries is precisely
- to relieve the programmer of dealing with the drudgery of the underlying
- machine. This is not a new development in computer science.
-
- >but being aware
- >of the limitations of the OS (I admit - it is a limitation) in this case
- >can, and does, lead to better programs.
-
- You are correct to an extent, but what happens when a new revision of the OS
- or a new chipset comes along? All of the assumptions you have made may
- change. This happened on the Amiga; all the Co|)3rZ that wrote self modifying
- code (justifiable on a 68000) found that it broke on a 68020. It is the 90's,
- for god's sake; applications should be written to be as portable and hardware
- independant as possible. It simply isn't worth making tinpot optimiztions
- anymore for a few % increase in speed.
-
- >This is certainly untrue in my experience.
-
- No, it is true. In a cooperative MT environment, there is no guarantee that
- a process will *ever* give up the CPU, let alone respond immediately to an
- event, without special hacks. Imagine you have a program that "yields" once
- every second. While this is running in the background, you will have an
- average response time of ~0.5s for your foreground tasks. Not good.
-
- >Faster than the X-windows interfaces we use at the uni,
-
- X is a whole different kettle of fish, designed to be network transparent and
- so on. This incurs a lot of overhead, but is worth it. As well as that, your
- home directory and various other things probably reside on a remote disk
- connected to your machine via NFS, which tends to reduce responsiveness a bit.
- Anyway, X on my P75 seems pretty responsive to me. Maybe I'm just slow ;-)
-
- >and faster than Windows 95 on my dads 486 DX2-66,
-
- No surprises there...
-
- >The PC has twice the MIPS of my Acorn - but the OS, combined with efficient
- >programming of applications means that the Acorn is much more responsive.
-
- Agreed. But Windows is shite anyway - you can't say something is good because
- it is better than something that is bad!
-
- >>It must have some CPU arbitration mechanisms then?
- >I'll agree with you - if only on the grounds that I don't know what you mean...
-
- I was wondering if there was anything "extra" in the MT model that RiscOS uses
- to ensure fairer task scheduling over and above that of cooperative MT.
-
- >A little out of context. I'm talking about a fast preview of a JPEG on a single
- >user machine here. For a fast response time - it makes sense to single task
- >for a second, and then let everything carry on...
-
- This is getting religious - I'm afraid we're going to have to agree to disagree
- on this point.
-
- --
- "I have trouble imagining death at that income level" - White Noise, D.Delillo
-
- Neil Clark, Transparent Telepresence Group
- <http://telepresence.dmem.strath.ac.uk/>
-