Hi all! I am running: Debian/unstable with libgl1-mesa-dri from experimental, ie 6.5.0.cvs.20060524-1 kernel 2.6.17-rc6-mm2 Xorg 7.0.0 Acer TravelMate 3012, (--) Chipset 945GM found and GL is extremely slow. The problem is verysimilar to bug #6672, but this time on i910 driver. Starting googleearth gives me: do_wait: drmWaitVBlank returned -1, IRQs don't seem to be working correctly. Try running with LIBGL_THROTTLE_REFRESH and LIBL_SYNC_REFRESH unset. and nothing is shown. Earlier I had dri/drm disabled and at this time googleearth was working better/at all. Any hints what I could try? Best wishes Norbert
Created attachment 5912 [details] [review] Check Pipe B for IRQ's Try applying this patch to your kernel sources in /usr/src/linux/drivers/char/drm/i915_irq.c
With this patch googleearth starts and works more or less as fast as in indirect rendering. So the one error is gone, the other that the rendering is quite slow is still present. Do you know any other simple program (pref. in Debian/unstable) with which I can test the performance? glxgears? But it does not give me any information (I thought it says something about blabla fps before). Thanks a lot and all the best Norbert
I'm not sure what's in Debian. You could checkout the Mesa source from CVS and use the sample programs in progs/demos to test performance. But if you are getting direct rendering, then it may just be that google earth is hitting a slower path that may not be optimized or possibly even a software path.
This patch fixes the googleearth issue on my Dell D410 laptop with an i915GM. The DFP on the laptop apparently is connected to PIPEB, normally. It would be nice if it made upstream sometime soon.
It's in the relevant git repos now. Closing.
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.