Path: utzoo!attcan!uunet!husc6!uwvax!tank!nic.MR.NET!xanth!ames!oliveb!amiga!kodiak
From: kodiak@amiga.UUCP (Robert R. Burns)
Newsgroups: comp.sys.amiga.tech
Subject: Re: LIGHT WEIGHT PROCESSES,  ALPHA VERSION
Message-ID: <3212@amiga.UUCP>
Date: 21 Dec 88 23:03:12 GMT
References: <8812180204.AA22260@postgres.Berkeley.EDU>
Reply-To: kodiak@tooter.UUCP (Robert Burns)
Organization: Commodore-Amiga Inc, Los Gatos CA
Lines: 23

In article <8812180204.AA22260@postgres.Berkeley.EDU> dillon@POSTGRES.BERKELEY.EDU (Matt Dillon) writes:
)
)	(2) EXEC appears to push the registers on the user stack when doing
)	    a context switch.  This means that the stack for each light
)	    weight process must be large enough to hold that.  Currently,
)	    the minimum stack allocated (specifying 0 in InitLWP()) is 68
)	    bytes.  Is this enough?  (Question ---->>>>> C-A).

Is this enough for what environment?  Are any ieee math libraries opened --
then you need room for the 68881 state.  Who else hangs off launch & switch?
Will 1.4 exec itself save floating point or MMU state information on the stack?

I understand why you want to see answers, not questions, but the last time
we told anyone stuff like "we guarantee the system will only use XXX bytes
for ..." was release 1.0.  That was for 256K systems.  You know the stories
of applications that shaved it too close and lost.

Be Prepared.  Better safe than sorry.  "Take this stack and pad it."  Etc.
-- 
Bob Burns, amiga!kodiak                   _
| /_  _|. _ |      Commodore __          |_) _ |_  _ )'
|<(_)(_)|(_\|<      /\ |  ||| _` /\      |_)(_\| )(_\ |
| \ Software    ___/..\|\/|||__|/..\___           Faith