Date: Thu, 29 Oct 1998 14:51:58 +0200 (EET)
From: Toni Rasanen <torasane AT mail DOT student DOT oulu DOT fi>
X-Sender: torasane AT paju DOT oulu DOT fi
To: djgpp AT delorie DOT com
Subject: Re: Banked vesa And another Question(as we already are here)
In-Reply-To: <3637F10D.6613@club-internet.fr>
Message-ID: <Pine.SGI.3.95.981029144629.1884A-100000@paju.oulu.fi>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=ISO-8859-1
Content-Transfer-Encoding: 8BIT
Reply-To: djgpp AT delorie DOT com

On Thu, 29 Oct 1998, Ludvig Larsson wrote:

> Hope someone can help me with a figure or two(I'v read some time
> ago that a bankswitch eats a houndred or so processor cycles?).

Really don't know the exact speed, but here are the facts;
BankedVBE1.2: for each bank (640x480x8 = 6banks) do: switch to rmode
  (real slow), call VESA interrupt (depends of how poor the 
  code is), and switch back to pmode (slow too).
BankedVBE2: for each bank do: call VESA service routine. Faster than
  previous, but still depends a lot of how good VESA routines are.
LFB: Just write to display memory all that is needed. No slow-down
  vesa-calls or such.

From my experience:
I myself have a P200 with ET6000-based VBE2-display. When I ran the
bank-switching mode in another P200 with S3-based VBE1.2, speed was
reduced by some 30-50% ... In my own computer the performance loss
when using banks is almost unnoticeable, though.

 ///           Toni Räsänen
///       torasane AT mail DOT student DOT oulu DOT fi