Subject: Re: Testers wanted for new TGA console code
To: Nathan J. Williams <nathanw@MIT.EDU>
From: R. C. Dowdeswell <>
List: port-alpha
Date: 04/07/2000 16:38:34
On 955144754 seconds since the Beginning of the UNIX epoch
Nathan J. Williams wrote:
>The code includes some accelerated text routines; tga_copyrows() gave
>about a factor-of-20 improvement over the native rasops scrolling
>(still kind of slow, though; "cat /usr/share/misc/termcap" clocked in
>at 15 seconds on my 500 MHz pc164). However, tga_putchar() only gave
>about a 30% improvement in non-scrolling character drawing
>operations. I had expected much more.. anyone care to speculate on
>the bottleneck?

Yep.  I've been thinking about this one for a while, but the problem
is that wscons only hands you one character at a time and hence
you can't get the full benefit of stippling 32 bits at a time, but
can only stipple 16 bits (or whatever the font width is.)

Also, on the copyrows(), we would do better to jump scroll a la
xterm, but that would require wscons to buffer input a bit using
a Nagley algorithm.  That would allow us to reduce the number of
calls to copyrows() significantly.

These improvements would increase the performance of wscons in
general for all consoles.

 == Roland Dowdeswell                      http://www.Imrryr.ORG/~elric/  ==
 == The Unofficial NetBSD Web Pages        http://www.Imrryr.ORG/NetBSD/  ==
 == The NetBSD Project                            http://www.NetBSD.ORG/  ==