www.delorie.com/archives/browse.cgi   search  
Mail Archives: djgpp/1993/04/13/11:25:31

Date: Tue, 13 Apr 93 10:38:10 EDT
From: DJ Delorie <dj AT ctron DOT com>
To: cs8023 AT mbox DOT ee DOT ncu DOT edu DOT tw
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: Re: Bad performance if we use EMM386?

> I feel very strange that why DJGPP use VCPI to swap in-out the memory?
> Why the sequence of memory usage is VCPI then XMS then other....?
> The performance for a 4MB ram machine use only XMS is faster than XMS +EMM386
> (or 386MAX or QEMM). Almost better than 200% ~ 300% !
> Is EMM386 will disable XMS or DJGPP's GO32 will confilt with EMM386 ?
> otherwise, why use EMS paging to allocate memory?
> please enlighten me :)

If VCPI is present, the only way to access the memory it controls is
though VCPI calls.  Since supporting multiple access strategies in one
program is more complex than only supporting one at a time, I decided
that if you're using VCPI then you'll only use VCPI.  Since most
programs can handle either Expanded or Extended memory, most
configurations will either have VCPI and give it all the memory, or
not have VCPI and leave all the memory as extended memory.  For most
users, this works out just fine, so I've never changed it.

If Extended (xms) memory is faster on your system, then use it.  Note
that this means that you can't use EMM386, QEMM, or any other EMS
emulator at all, as that would require VCPI.

DJ

- Raw text -


  webmaster     delorie software   privacy  
  Copyright © 2019   by DJ Delorie     Updated Jul 2019