[comp.sys.amiga] Multiprocessing-Multitasking

DAVEA%CERNVM.BITNET@cunyvm.cuny.edu (David Almond) (04/10/89)

> In that the amiga has dedicated processers for graphics etc it is, to some
> degree multitasking, but not in any sophisticated, scheduled sense.

>>>Do you write for BYTE, by any chance?  That's dead wrong, on every account.

>>>The Amiga is FULLY multitasking, in the sense that the processor is time
>>>sliced among many processes.  Much in the same vein as UNIX, OS/9, Aegis,
>>>VMS, etc.

  No, if I did the sales would soar. What I meant to say, and not wishing
to burn precious space, is that the though the Amiga is multitaskingfor
all the reasions you outlined is not fully multitasking beause there
is no scheduler which controls overall running of tasks. So that it
would no how much spare capacity exists on each intelligent processing
device, as it processes, and thus be able to dispatch the next processing
task in the most efficent manner.


  Now I know the Amiga is good, but its not that good .... nes pa


  ....... Dave

martens@ketch.cis.ohio-state.edu (Jeff Martens) (04/10/89)

In article <12700@louie.udel.EDU> DAVEA%CERNVM.BITNET@cunyvm.cuny.edu (David Almond) writes:

:: In that the amiga has dedicated processers for graphics etc it is, to some
:: degree multitasking, but not in any sophisticated, scheduled sense.

::::Do you write for BYTE, by any chance?  That's dead wrong, on every account.

::::The Amiga is FULLY multitasking, in the sense that the processor is time
::::sliced among many processes.  Much in the same vein as UNIX, OS/9, Aegis,
::::VMS, etc.

:  No, if I did the sales would soar. What I meant to say, and not wishing
:to burn precious space, is that the though the Amiga is multitaskingfor
:all the reasions you outlined is not fully multitasking beause there
:is no scheduler which controls overall running of tasks. So that it
:would no how much spare capacity exists on each intelligent processing
:device, as it processes, and thus be able to dispatch the next processing
:task in the most efficent manner.

You're confused.  For the reasons the other poster outlined, the Amiga
does have a full-blown scheduler:  tasks are started, stopped,
time-sliced, and prioritized.  What else do you want from a scheduler?
Certainly the Amiga scheduler "controls the overall" scheduling "of
tasks." 					     ----------

I'm not sure what you meant by "spare capacity...", but most OS
schedulers aren't brilliant, and don't do much more.  You may be
thinking of resource tracking or protection, which the Amiga does
lack, but these wouldn't be part of the scheduler anyhow.
-=-
-- Jeff (martens@cis.ohio-state.edu)

...and on Wall St., the Tao is unchanged in moderate trading...

elg@killer.Dallas.TX.US (Eric Green) (04/12/89)

in article <12700@louie.udel.EDU>, DAVEA%CERNVM.BITNET@cunyvm.cuny.edu (David Almond) says:
> to burn precious space, is that the though the Amiga is multitaskingfor
> all the reasions you outlined is not fully multitasking beause there
> is no scheduler which controls overall running of tasks. So that it
> would no how much spare capacity exists on each intelligent processing
> device, as it processes, and thus be able to dispatch the next processing
> task in the most efficent manner.

Say WHAAA??? are you SURE that you don't work for BYTE??? (Yes, I know
that you denied it -- but this is typical of Byte misinformation).

The Amiga DOES have a central task scheduler -- what do you think Exec
is? Now, there's a few problems with the Amiga's task scheduler -- it
allows high priority processes to hog all of the CPU without dribbling
any down to low priority processes, in the event that the programmver
of the high priority process puts a wait loop instead of a Wait() call.
But that is irrelevant to the average user (who runs all of his
programs at priority 0, where un-blocked tasks are timesliced
equally), and perfect for real-time work (which is what Exec was
designed for in the first place).

So, while the Amiga's task scheduler isn't of the level of, say, the
Multics task scheduler (which took into account size of program, how
many keys you typed in the last few minutes, and lots of other
esoteric info -- I kid you not), it DOES have one. 

Or perhaps you wanted to say "let other processes run while you're
using the blitter"? Look at the QBlit () routine in graphics.library,
which allows you to queue blitter requests. Or WaitBlit() which puts
your process to sleep while your blitter request is going (thus
letting other programs run). Or.... well, you get the picture.

Or perhaps you wanted to say, "have the disk handler queue requests &
sort them so that the disk drive doesn't gronk"? That has nothing to
do with the multitasking ability of the Amiga -- just with a very
badly designed disk handler (:-().

The Amiga is multitasking. Period. It's the multi-processing part
that's only "semi" (the blitter and copper aren't "real" processors,
but they "sort of" count). And while the internals of the process
scheduler would require work for a multi-processor Amiga, a
shared-memory multi-processor AmigaDOS would work just fine (just as
long as you have write-through caches, & bus-sniffing). 

--
|    // Eric Lee Green              P.O. Box 92191, Lafayette, LA 70509     |
|   //  ..!{ames,decwrl,mit-eddie,osu-cis}!killer!elg     (318)989-9849     |
| \X/            Amiga.  The homestation for the blessed of us.             |

keithh@atreus.uucp (Keith Hanlan) (04/13/89)

In article <42406@tut.cis.ohio-state.edu> Jeff Martens <martens@cis.ohio-state.edu> writes:
>In article <12700@louie.udel.EDU> DAVEA%CERNVM.BITNET@cunyvm.cuny.edu (David Almond) writes:
>
>:: In that the amiga has dedicated processers for graphics etc it is, to some
>:: degree multitasking, but not in any sophisticated, scheduled sense.
>
>::::The Amiga is FULLY multitasking, in the sense that the processor is time
>::::sliced among many processes.  Much in the same vein as UNIX, OS/9, Aegis,
>::::VMS, etc.
>
>:to burn precious space, is that the though the Amiga is multitaskingfor
>:all the reasions you outlined is not fully multitasking beause there
>:is no scheduler which controls overall running of tasks. So that it
>:would no how much spare capacity exists on each intelligent processing
>:device, as it processes, and thus be able to dispatch the next processing
>:task in the most efficent manner.


	I think the confusion is that Mr. Martens is talking about
	*multi-processing* but calling it multi-tasking.

	Yes the amiga is a multi-processing computer in that it has multiple
	processors but it is not multi-processing in the same sense as a
	multi-CPU machine. This is hardly surprising. The co-processors are
	dedicated function chips.

	Does this clear things up?
	Keith Hanlan		{uunet!attcan!}utgpu!bnr-vpa!bnr-fos!atreus!keithh
	Bell-Northern Research
	Ottawa, Canada

martens@cis.ohio-state.edu (Jeff Martens) (04/14/89)

In article <397@bnr-fos.UUCP> Keith Hanlan writes:
>In article <42406@tut.cis.ohio-state.edu> Jeff Martens <martens@cis.ohio-state.edu> writes:
:>In article <12700@louie.udel.EDU> DAVEA%CERNVM.BITNET@cunyvm.cuny.edu (David Almond) writes:
:>
:>:: In that the amiga has dedicated processers for graphics etc it is, to some
:>:: degree multitasking, but not in any sophisticated, scheduled sense.
:>
:>::::The Amiga is FULLY multitasking, in the sense that the processor is time
:>::::sliced among many processes.  Much in the same vein as UNIX, OS/9, Aegis,
:>::::VMS, etc.
:>
:>:to burn precious space, is that the though the Amiga is multitaskingfor
:>:all the reasions you outlined is not fully multitasking beause there
:>:is no scheduler which controls overall running of tasks. So that it
:>:would no how much spare capacity exists on each intelligent processing
:>:device, as it processes, and thus be able to dispatch the next processing
:>:task in the most efficent manner.


>	I think the confusion is that Mr. Martens is talking about
>	*multi-processing* but calling it multi-tasking.

>	Yes the amiga is a multi-processing computer in that it has multiple
>	processors but it is not multi-processing in the same sense as a
>	multi-CPU machine. This is hardly surprising. The co-processors are
>	dedicated function chips.

>	Does this clear things up?
>	Keith Hanlan		{uunet!attcan!}utgpu!bnr-vpa!bnr-fos!atreus!keithh
>	Bell-Northern Research
>	Ottawa, Canada


No, this muddies it up.  Although none of my posting was repeated (so
I think you were attributing to me what soemone else said), I was
talking about multasking, as was the person quoted above who pointed
out that Unix, VMS, etc. multitask in the same sense as the Amiga (and
vice-versa).  Multiprocessing is the use of multiple processors; the
Amiga does this in a very asymmetric sense since there are processors
inside dedicated to screen handling and the like.  Symmetric
multiprocessing would require more than one 68000, which my Amiga
doesn't have.  Multitasking, on the other hand, just means there can
be more than one thread of control at a time.  See any intro undergrad
OS text, like Deitel's.

My main objection was to the person who claimed that the Amiga doesn't
have a scheduler; this is clearly false.

On a second read through keithh's posting, perhaps he meant that
the person I was responding to (the person quoted above as claiming
the Amiga isn't fully multitasking) was confused between multitasking
and multiprocessing.  If so, maybe he should just pay more attention
to who said what.
-- 
-- Jeff (martens@cis.ohio-state.edu)

....and on Wall St., the Tao is unchanged in moderate trading...

kim@beowulf.ucsd.edu (Geoffrey K Kim) (04/14/89)

In article <397@bnr-fos.UUCP> keithh@atreus.UUCP (Keith Hanlan) writes:
>
>	I think the confusion is that Mr. Martens is talking about
>	*multi-processing* but calling it multi-tasking.
>

Unfortunately I think this confusion was brought about by certain 
computer and software manufacturers that ended up *bastardizing* the
term 'multi-tasking'.  PROVIDING A MANUAL TASK SWITCHING CAPABILITY
IS *NOT* TRUE MULTI-TASKING!  Sorry, I just had to get that off my
chest.  Now I feel better, thank you.

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| kim @beowulf.UCSD.EDU (Home of the Garden Weasles)                  |
|        "... ENGAGE!" -- Jean Luc Picard, STTNG                      |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

rwallace@vax1.tcd.ie (04/21/89)

In article <7819@killer.Dallas.TX.US>, elg@killer.Dallas.TX.US (Eric Green) writes:
> in article <12700@louie.udel.EDU>, DAVEA%CERNVM.BITNET@cunyvm.cuny.edu (David Almond) says:
>> to burn precious space, is that the though the Amiga is multitaskingfor
>> all the reasions you outlined is not fully multitasking beause there
>> is no scheduler which controls overall running of tasks. So that it
>> would no how much spare capacity exists on each intelligent processing
>> device, as it processes, and thus be able to dispatch the next processing
>> task in the most efficent manner.
> 
> Say WHAAA??? are you SURE that you don't work for BYTE??? (Yes, I know
> that you denied it -- but this is typical of Byte misinformation).
> 
> The Amiga DOES have a central task scheduler -- what do you think Exec
> is?

Of course the super-smart mainframe OS task schedulers which can share out
.001 MIPS of processing power more fairly among tasks will gobble up a hundred
times that much processing power because they're so inefficiently written.
At least on the Amiga almost all the processor time actually goes to the tasks
not the multitasking system.

"To summarize the summary of the summary: people are a problem"
Russell Wallace, Trinity College, Dublin
rwallace@vax1.tcd.ie