Francesco Poli wrote: > On Fri, 25 Sep 2009 22:28:06 +0200 Brice Goglin wrote: > > >> Francesco Poli wrote: >> >>> Nonetheless, the test was not completely error free. Indeed, some >>> worrying entries were logged to /var/log/kern.log: >>> >>> >>> [ 582.975686] Xorg:5396 conflicting memory types 90000000-95000000 >>> uncached-minus<->write-combining >>> > [...] > >>> As a consequence, I am *not* sure that the bug can be safely considered >>> as fixed... >>> >>> >> This looks like a kernel problem to me. It's probably the Xserver asking >> the kernel to set some memory range as WC but the kernel PAT code fails >> to do so for some reason, it sets UC- minus instead (things should be >> slower then). >> > > Things are veeeryyyy slooooowwww, indeed! > It seems that the long-awaited xorg migration to testing caused my > already slow video chipset to become even slower with OpenGL > applications... :-( >
Adding a WC MTRR manually might help. But you may have to boot with "nopat" on the kernel command-line first since the WC MTRR is ignored if the kernel sets UC- with PAT IIRC. Brice -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
