From: Charles Sandmann <sandmann AT new-orleans DOT NeoSoft DOT com>
Subject: Re: V2
To: UCKO AT VAX1 DOT ROCKHURST DOT EDU (Aaron Ucko)
Date: Fri, 27 Jan 1995 11:45:58 -0600 (CST)
Cc: palomino!edward AT uunet DOT uu DOT net, djgpp AT sun DOT soe DOT clarkson DOT edu

> words with a grain of salt, but I believe the biggest problem deals
> with virtual memory mapping.  Under VCPI, it is possible to guarantee
> that physical memory gets mapped to E0000000--EFFFFFFF, etc.  Under
> DPMI, this is no longer possible, which is a problem to several bits
> of code, especially those pertaining to graphics.  

The 0xe00xxxxx area isn't a problem at all since the farptr functions
and mem routines make that one trivial.  

The virtualized linear buffer at 0xd00xxxxxx is a problem, but this
is only for hi-res graphics writing directly.  That code will need to
be changed to be bank-switching aware.  Calling grx routines to do the
work for you should not be a problem, that works.

> BTW, have any GRX
> drivers been made supporting direct linear mapping (I know that at
> least my Cirrus Logic card supports this)?  If so, the problem would
> be somewhat alleviated...

There are very few drivers at all for GRX 2.0 right now, but for 
advanced cards that allow this mapping have been considered.