[comp.sys.mac.programmer] LSC 3.0 memory...it's not their fault!

hugo@maxlin.dartmouth.edu (Peter Su) (06/23/88)

Hi folks...

Well, with all the moaning about the memory requirements of LSC and
other things, I thought I would try to say something that's been on my
mind...se, I'm not really mad at THINK, or the Fullwrite people,
because it's not really their fault that their software needs so much
memory to run with Multifinder, in my mind, the problem is Apple's
non-OS.  

Let me explain:  See, on some computers, when you run multiple tasks,
the OS keeps can (gasp!) swap out code and data that are not currently
being referenced, and (gee!) you don't have to allocate like and
entire MB of memory for (say) Excel, when in fact you may actually use
only half of it at any given time, but there is some extreme case where
it will need it all.  See, the OS manages the DYNAMIC memory
requirements of the programs that are running, and so the developer
doesn't have to worry so much.

This business with paritions is just such an incredibly dirty hack
that it's almost sickening.

So when you are moaning about needing 2MB for LSC, reember that Apple
is also to blame, because their OS, well, isn't one.

Disclaimer:

Now, you all are thining, geez, what a UNIX/MSDOS/IBM bigot.  Not so!
I think the mac is great, but I can't stand the thought of programming
the thing,because it is just too painful.

Cheers,
Pete

adail@pnet06.cts.com (Alan Dail) (06/23/88)

One thing I think you are overlooking in complaining about the fact that
the Mac OS doesn't support virtural memory yet is that any system I know
of that supports virtural memory seem to require 2-4 MB just for the OS.

Alan Dail

UUCP: {crash uunet}!pnet06!adail
ARPA: crash!pnet06!adail@nosc.mil
INET: adail@pnet06.cts.com

blknowle@uokmax.UUCP (Bradford L Knowles) (06/24/88)

In article <8980@dartvax.Dartmouth.EDU> hugo@maxlin (Peter Su) writes:
>[... stuff deleted ... ] the problem is Apple's
>non-OS.  
>
>Let me explain:  See, on some computers, when you run multiple tasks,
>the OS keeps can (gasp!) swap out code and data that are not currently
>being referenced, and (gee!) you don't have to allocate like and
>entire MB of memory for (say) Excel, when in fact you may actually use
>only half of it at any given time, but there is some extreme case where
>it will need it all.  See, the OS manages the DYNAMIC memory
>requirements of the programs that are running, and so the developer
>doesn't have to worry so much.
>
>Cheers,
>Pete

Pete, the system already does this.  If a resource is listed as "purgeable"
then the system will "unload" it if more memory is required.  Admittedly
MultiFinder doesn't dynamically reallocate memory for programs, but this
is one of its first incarnations -- when Apple does add this feature, then
you can be guaranteed that people will bitch because the system takes up
too much meomory, a price we will all have to pay if we want mainframe
and minicomputer performance out of a microcomputer.

So waddya want?  The universe on a dime budget?  Not even Apple can
deliver *THAT*!

-Brad Knowles

UUCP: ...!ihnp4!occrsh!uokmax!blknowle     ARPA: blknowle@uokmax.ecn.uoknor.edu
SNAIL: 1013 Mobile Circle
       Norman, OK  73071-2522
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Philosophy of Boris Yeltsin: "If one wants to be unemployed, one will
	criticize ones' boss.  If one wants to be sent to Siberia, one
	will criticize the wife of ones' boss."
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Disclaimer: (The above opinions are my own.  They have nothing to do with the
		Univeristy of Oklahoma nor intelligance of any sort. :-)

jimc@iscuva.ISCS.COM (Jim Cathey) (06/25/88)

In article <242@hodge.UUCP> adail@pnet06.cts.com (Alan Dail) writes:
>One thing I think you are overlooking in complaining about the fact that
>the Mac OS doesn't support virtural memory yet is that any system I know
>of that supports virtural memory seem to require 2-4 MB just for the OS.

The Kernel (both code and data) of our Unix variant occupies about
250K.  We run in 1MB of physical RAM, and can in fact run (though quite
a bit slower due to increased swapping) in 512K of RAM.  The virtual
space is 2MB right now because of a limit in the MMU.

It _is_ possible to run virtual on smaller machines.  The biggest
problem seems to be programmers who then don't realize that these _are_
smaller machines!

Of course, I am not proposing that VM could just drop into the Mac OS for a
paltry memory fee.  There are a lot of technical issues to solve (unless you
want VM with no intertask protection -- that would be quite a bit easier).

+----------------+
! II      CCCCCC !  Jim Cathey
! II  SSSSCC     !  ISC Systems Corp.
! II      CC     !  TAF-C8;  Spokane, WA  99220
! IISSSS  CC     !  UUCP: uunet!iscuva!jimc
! II      CCCCCC !  (509) 927-5757
+----------------+
			"With excitement like this, who is needing enemas?"

jmunkki@santra.HUT.FI (Juri Munkki) (07/01/88)

In article <1658@iscuva.ISCS.COM> jimc@iscuva.ISCS.COM (Jim Cathey) writes:
>Of course, I am not proposing that VM could just drop into the Mac OS for a
>paltry memory fee.  There are a lot of technical issues to solve (unless you
>want VM with no intertask protection -- that would be quite a bit easier).

Of course that is what I want and what I think most people would like. Since
the Mac currently has absolutely no protection between applications, there
would be no harm from a virtual system without protection. I estimate that it
would take a good Mac programmer about a week to write a beta version and
about two months to make it a product.

Tell me why it can't be done. It wouldn't hurt apple's image to be able to
advertise that the Mac has virtual memory.

I already suggested this earlier and got no replies. Are MMUs harder to find
than 1M DRAMs?

Juri Munkki
jmunkki@santra.hut.fi
jmunkki@fingate.bitnet

P.S. Give me a MMU and the manuals and I'll write a swapper.

landman%hanami@Sun.COM (Howard A. Landman) (07/07/88)

In article <242@hodge.UUCP> adail@pnet06.cts.com (Alan Dail) writes:
>One thing I think you are overlooking in complaining about the fact that
>the Mac OS doesn't support virtural memory yet is that any system I know
>of that supports virtural memory seem to require 2-4 MB just for the OS.

Can't you buy a uVAX II running Ultrix with only 1 MB?  More is recommended,
but doesn't 1 MB work?

Certainly, Sun 2s were available in 2 MB configurations, and could run
reasonably large programs.  Here are the memory configurations that I have
been able to discover for the Sun family (disclaimer: I haven't been here
long enough to know what I'm talking about!):

Model		Min-Max Mem (MB)
4/110		8-32
386i/250	8-16
386i/150	4-16
3/200		8-128
3/180,/160	4-16
3/140,/110	4-12
3/60		4-24
3/50		4
2/160,/130,/50	2-?

Roughly, the minimum memory configuration seems to double every generation.
Anyway, a 2 MB system puts the lie to "require 2-4 MB just for the OS".

	Howard A. Landman
	landman@hanami.sun.com
	UUCP: sun!hanami!landman

dcw@ecsvax.uncecs.edu (Douglas C. Wilson) (07/08/88)

In article <59085@sun.uucp> landman@sun.UUCP (Howard A. Landman) writes:
>Can't you buy a uVAX II running Ultrix with only 1 MB?  More is recommended,
>but doesn't 1 MB work?
>	Howard A. Landman
>	landman@hanami.sun.com
>	UUCP: sun!hanami!landman
The current version of ULTRIX-32 requires a minimum of 4 meg just to boot.
Any real work takes more.
-- 
Doug Wilson				Internet: dcw@rwlvax.chem.unc.edu
Dept. of Chemistry			BITNET	: WILSON@UNCCHEM,WILSON@UNCVAX
Univ. of North Carolina			UUCP	: mcnc!ecsvax!dcw
Chapel Hill, NC  27599			Phone	: 919-962-1582, 1583

phssra@emory.uucp (Scott R. Anderson) (07/12/88)

In article <59085@sun.uucp> landman@sun.UUCP (Howard A. Landman) writes:
>In article <242@hodge.UUCP> adail@pnet06.cts.com (Alan Dail) writes:
>>One thing I think you are overlooking in complaining about the fact that
>>the Mac OS doesn't support virtural memory yet is that any system I know
>>of that supports virtural memory seem to require 2-4 MB just for the OS.
>
>Certainly, Sun 2s were available in 2 MB configurations, and could run
>reasonably large programs.

The Sun-1s came with 1MB standard.  Running SunWindows was painful, but
otherwise it worked fine (if your programs weren't monsters).  As I recall the
kernel could be as small as 400K.  This might still be true, but it's been a
while since I adminstered one, so I don't know for sure.  The large amounts
of memory currently seen on Suns are the result of lower memory prices and the
realization that SunView runs much, much better if you never have to swap :-).

*                                     Scott Robert Anderson
  *      **                           gatech!emoryu1!phssra
   *   *    *    **                   phssra@emoryu1.{bitnet,csnet}
    * *      * *    * **
     *        *      *  * * * * * * * * * * * * * * * * * * * * * * * * * * * *