You probably heard the screams of terror at the prospect of adding 50k lines of broken code. The result of that shock was that it did produce a viable alternative: use a shadow fb and avoid dynamic reallocation and associated incoherency issues.
The benefits of this approach is that we get KMS (hotplug, resizable framebuffers, rotation, reliable suspend and resume), overlays (video) reasonable performance, and continued development. You lose GL acceleration. (But using GEM it would suffer the same incoherency problems that plagued X, and unless you were also using an old DRI1 compatible version of Mesa, legacy/ums could do no better.) You also lose EXA acceleration, but that turns out to be a good thing in the general case, but synthetic benchmarks will show decreased performance. The branch is available for testing from http://cgit.freedesktop.org/~ickle/xf86-video-intel/log/?h=shadow To enable the use of the shadow, you will have to add Section "Device" Option "Shadow" "True" EndSection to your xorg.conf. -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
