rx__: hm.. new fglrx to play with
z3ro: rx__: the ubuntu one?
rx__: november release
z3ro: hmm ok. I guess that is newer than 8.543 (just me, or is the version numbering really inconsistant)
rx__: it's inconsistent :)
rx__: 8.11 = 8.552
rx__: ubuntu = 8.543
rx__: 8.10 = 8.542
z3ro: hmm ok. I'll have to give 8.11 a try later...
z3ro: I'm going to try intercepting the ioctls so might as well do it on the latest version.
rx__: there's xorg 1.5 support thrown in
rx__: and i'm assuming 2.6.28 support too
rx__: if you don't get it to work.. let me know :)
rx__: is happy to dump
z3ro: well, it's more that I'm writing a new dumper. :)
z3ro: I'm not entirely happy with the approach I'm currently using in revenge. it's possible some packets get lost.
spstarr: looks at the garbled flash video
z3ro: ok, so don't delay the ioctl too long...
z3ro: airlied: dumping the ioctls doesn't give anything looking like CP packets...
z3ro: I think maybe they are just passing pointers to packets in the ioctls... but I don't see a clear structure yet.
z3ro: looks like they are passing a structure via the ioctl that holds a lot more than just a pointer to the packets.
z3ro: yay for diff! :)
z3ro: well, this is possible, but I think it's going to be easier to just trace userspace MMIO from within the kernel.
z3ro: I want to see how that works out before moving forward... there is going to be quite a bit of guess work in the ioctl approach.
Remosi: z3ro, what are you trying to figure out? :)
rx__: figuring out what fglrx does
spstarr_coding: hmm no DRI2 for me tonight ;/
spstarr: 101 is broken with nomodeset and 1xx kernels fix it.. but not ready built
spstarr: 109 has hit the world
spstarr: composite is fast... very fast...
spstarr: full screen firefox minimize maximize with kwin gene effect, there is no stalls
spstarr: i have blocked X from VT switching
spstarr: airlied: donno what drm bits you changed but .. wow...
spstarr: ive not see r3xx this fast
airlied: I took out the sleep 1's :)
spstarr: ha :)
spstarr: X is still eating CPU but this all looks to be accel
spstarr: it's good enough
spstarr: no flash corruption with composite on
spstarr: i noticed gdm's initial wallpaper is not corrupt (one corner showing) either now
spstarr: there is pixmap corruption on GTK menus on load
spstarr: no Qt menus
spstarr: tries to capture
spstarr: airlied: maybe a small video capture to show the effect