Sender: |
|
Date: |
Mon, 29 Mar 1999 14:10:52 -0800 |
Reply-To: |
|
Content-type: |
text/plain; charset=US-ASCII |
Subject: |
|
From: |
|
Content-transfer-encoding: |
7BIT |
In-Reply-To: |
|
Organization: |
General Magic |
MIME-Version: |
1.0 |
Parts/Attachments: |
|
|
On 28 Mar 99, at 15:36, Max Timchenko wrote:
> How exactly can 2 cards live in one system? they compete for same
> address space ($a000:0 and up), won't it cause conflicts?
All VGA cards have to provide modes where the video RAM can be mapped
into the A000 segment (64K at a time) in order to be accessible from
DOS programs. That is true.
However, constantly mapping and re-mapping this address space is
ugly, time-sonsuming, and pretty boring. Most modern video cards
*also* map their entire RAM to space at a high (non-swappable) virtual
address, where protected-mode software can read and write it directly;
the drivers for the card know where this mapping is, or how to find it.
So there's only a conflict at A000 if both cards are trying to
provide real-mode (DOS-compatible) access to video RAM. As long as at
least one of them has this mapping suppressed, there's no conflict.
[Actually, under NT and 9x, this mapping is generally done by the OS,
and used to connect to a "DOS window" on the screen, instead of
connecting directly to the video RAM. So it's "extremely normal" for a
video card under Win32 not to map to A000 at all.
[There is a secondary issue that two VGA/SVGA cards would normally
want to use the same range of I/O port numbers and IRQ (if any);
solving that part of the problem seems like a natural extension of PnP,
which would I think be why this feature is only supported by fairly
recent cards.]
David G
PCBUILD maintains many useful files for download
on our web site - visit our download page at:
http://nospin.com/pc/files.html
|
|
|