Xref: news-dnh.mv.net comp.os.msdos.djgpp:136
Path: news-dnh.mv.net!mv!news.sprintlink.net!gatech!swrinde!pipex!sunic!sunic.sunet.se!news.uni-c.dk!diku.dk!terra
From: terra AT diku DOT dk (Morten Welinder)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: fork()/join()
Date: 5 Jun 1995 17:38:21 GMT
Organization: Department of Computer Science, U of Copenhagen
Lines: 27
Sender: terra AT tyr DOT diku DOT dk
References: <3qlh05$kcn AT odin DOT diku DOT dk> <3qv6e7$h59 AT jabba DOT ess DOT harris DOT com>
Nntp-Posting-Host: odin.diku.dk
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Dj-Gateway: from newsgroup comp.os.msdos.djgpp

azimmer AT rsa DOT hisd DOT harris DOT com (Alan D Zimmerman) writes:

>> fork(), btw, is going to be a pig since it will have to copy the
>> entire running process.  The multi-tasking exec's will behave much
>> better.

>Why?  I have written programs in Ada using Meridian OpenAda DOS that 
>supports tasking and runs just fine on an 8086/8088, and 80286 under
>DOS.  I know this is a DJGPP (which supports GNAT) newsgroup and not
>an Ada newsgroup, but your comments are not necessarily true, it doesn't
> have to run like a pig.


I don't know the relavant primitives of ADA, but with fork()
the semantics is such that each branch gets it own copy of
all memory, (specified) file handles, and a zillion other
things.  There a big difference here between multi-threading
and multi-tasking.  I think ADA only supports the first.

The reason it is not a pig under your normal unix is that
there you can do copy-on-write for memory, so most of the
memory will not actually get copied.

We can't do copy-on-write on DPMI 0.9; Some DPMI 1.0 hosts
may (it's optional) support it.

Morten