On Thu, 2001-11-29 at 19:12, Derrik Pates wrote:
> I have an iBook running Linux, with BenH's 2.4.16-ben0 kernel, and=20
> Debian XFree packages (with one minor hack to get line width right). The=20
> UniNorth AGPGART seems to be working, and I'm able to successfully do=20
> DRI (all the GL XScreenSaver hacks work, glxgears works, glxinfo says=20
> direct rendering is enabled).
agpgart is really stable for you? Most things like glxgears work fine on
my Pismo with it, but a few other things like tuxracer reliably lock up.
> However, some of the XScreenSaver GL hacks, when they end, leave the mous=
> pointer in a weird state - as an (approx.) 64x64 box, with what looks
> like ~24 lines inverting, ~24 lines transparent, then ~8 lines
Never seen that. Or heard of it, for that matter. :)
Sounds like the hardware cursor image gets overwritten, maybe an
offscreen memory glitch or related to agpgart?
> Other than that, it works quite well, though the frame rate seems a
> bit on the low side (~190 fps according to glxgears).
That's too low indeed. I get ~350 without agpgart and almost 500 with
it. (Or is your CPU slower than 300 MHz?)
Are you running a 'complex' window manager like Enlightenment or any 2D
clients at the same time? If yes, the context switching overhead might
explain this. That will be better with current CVS and 4.2.0.
> Also, I'm still seeing the bleedthrough effect that had been mentioned on
> this list previously with Rage128 DRI support - is there a fix in CVS
> that eliminates this problem? Maybe lined up for the upcoming 4.2.0
> release? :)
You mean the 'meltdown' effect, which is a symptom of the chip hanging
up? Yep, that might be better too. Sounds like you should try current
CVS, eh? :)
Earthling Michel D=E4nzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member / CS student, Free Software enthusiast