I've got a problem with a PCI Radeon on an Alpha that I've just about
started tearing my hair out over, and I don't have that much to
waste, hence my plea for help... ;-}
Background: ported the pcigart-0-0-1-branch to Alpha and successfully
ran OpenGL apps (gears, of course, and tuxracer :-) on ATI Rage 128
PCI. Took a snapshot late last week of pcigart-1-0-0-branch,
finishing the port from 0-0-1, then made my Alpha fixes, and Rage 128
worked. Moved on to the Radeon code, and got it working over the
weekend.
Status: Radeon code seems stable (16 and 32 bit depths), runs all OGL
apps thrown at it, but...
Problems:
0. have to turn off XAA offscreen pixmaps even non-direct-rendering,
because of screen corruption.
1. code detects 48MB video RAM, while the card only has 32MB; worked
around via XF86Config override, haven't investigated fully.
2. video and/or rendering problem (uncertain, but it looks like both).
Things I've tried or observed:
1. setting "noaccel" or running 8/15 bit depths, video/rendering fine.
2. problem happens at both 16 and 32 bit depths.
3. move a black xterm even a small amount, and the black body fills
with "snow" gradually but quickly; other windows show similar effect.
4. re: #3, turning off XAA screen-to-screen does NOT fix the problem
5. there are vertical stripes, looking like 1-2 pixels wide, every
32/64/128 pixels, with every other horiz row OK (look kinda like
dotted vertical lines).
6. cursor shape is OK, but there's a small banded rectangle that follows
it around, looking like the cursor data/mask is bad somehow
7. fonts are all a bit mishapen.
8. console (VGA) textmode font rendering is messed up after exiting X.
running X non-direct-rendering and exiting cleans it up.
So, any ideas? What can I try next? What other info do you need?
Thanks for any help.
BTW, real nice job on the PCIGART support! I really only had to muck
with the PAGE_SIZE difference between Alpha and Intel in setting up
the PCIGART page table.
Also, the virt-to_bus() usage will need replacing for handling the
job on machines with more memory than can be addressed directly. That's
my next task after the Radeon problem is solved...
Thanks again for all your work.
--Jay++
-----------------------------------------------------------------------------
Jay A Estabrook Alpha Engineering - LINUX Project
Compaq Computer Corp. - MRO1-2/K20 (508) 467-2080
200 Forest Street, Marlboro MA 01752 [EMAIL PROTECTED]
-----------------------------------------------------------------------------
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/dri-devel
