Created attachment 44338 [details] The xorg.conf used for the test In my attempts to fix my performance issues with my Radeon card, I did a test run of the X server using cmdline "valgrind --tool=callgrind /usr/bin/X.nosuid" to see, where the X server burns all those CPU cycles I see in top. The answer: In libexa.so: exa_unaccel.c. That's obvious, because if it were accellerated, I wouldn't have made the test run in the first place, what's not obvious: I explicitly asked the X server to use XAA (to test, whether that's better). It seems the X server is just ignoring that particular option alltogether.
Created attachment 44339 [details] The Xorg.log file generated by the run
"AccelMethod" is a driver option. The radeon driver can't support XAA with KMS.
Oh, didn't know (found many references to the options in forums and just tried them ...). How about writing a warning message in Xorg.log like "XAA is not supported when KMS is enabled - falling back to EXA"?
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.