"Brian S. Julin" wrote:
> 
> On Fri, 20 Apr 2001, Sven Garbade wrote:
> > oh, were can I get the sources? Sorry, but I haven't found them on
> > www.ggi-project.org. And another question: is it possible to sync a ggi
> > application with the vertical retrace? I've tried several ways with
> > medium success (e.g. using SDL and reading the VGA vrt bit).
> 
> LibGGIMisc (the one in the GGI snapshots right now) will
> support vrt sync on the svgalib target.  It basically just
> does the same thing, though.

The problem with SDL is, that you have to call a special update function
to update the changed areas on screen. This function doesn't care about
"self-designed" vrt functions, nevertheless, you get better results. 

> I have a new proposed version of LibGGIMisc released, and another one
> to follow (the next release will require LibGalloc).  I would
> expect that that will be ready for production use in about a month.
> That version adds vrt sync using the fbdev target, which will use
> the more advanced chipset-specific registers.  However the only
> fbdev driver in the kernel that actually implements vrt is matroxfb,
> so unless you want to add that function to the fbdev driver for
> your chipset, it won't be useful.

I´ve got a Matrox G400, so matroxfb should work.

> I also have some code where I have been playing around with the
> standard VGA VRT bits and the pentium-class rdtsc assembler instruction,
> to try to figure out where the raster beam is just by reading the
> processor time-stamp counter, which does not require any special
> access privilages or bus IO.  If you are into kernel hacking, then
> I can send you that code (will need updating for 2.4, last time it
> was updated was late 2.2/early 2.3)  It is kind of flaky right
> now, as I need a better algorithm to quickly calibrate the equation
> for finding the relationship between the retrace and the time-stamp
> counter.

Actually I'm still using kernel 2.2.18 (with RTAI-Linux). It would be
nice, if you can send me the code.

Many Thanks, Sven

Reply via email to