On Mon, Jun 3, 2013 at 2:52 PM, Daniel Drake <d...@laptop.org> wrote: > Just gave the latest git driver a quick spin on OLPC XO-1.5 (VX855) > now that EXA has been enabled again. > > However, composite doesn't seem to be working. viaExaCheckComposite_H6 > is called a lot, and returns TRUE in most cases, but the > PrepareComposite and Composite functions are never called (tested by > inserting ErrorF statements, nothing gets printed). > > Looking a little deeper, I see that exaTryDriverComposite calls > exaGetOffscreenPixmap() on the pixmap it is working with, which calls > down into exaPixmapHasGpuCopy_classic(), which returns: > > ret = (pExaPixmap->use_gpu_copy && pExaPixmap->fb_ptr); > > In this case, use_gpu_copy and fb_ptr always seem to be FALSE and NULL.
I dug further. This is because VIAPreInit() forces the greedy migration heuristic, which effectively disables composite and anything involving offscreen pixmaps. That still doesn't fix composite on XO-1.5 though. Now we get failures in PrepareComposite because viaExaIsOffscreen() always returns FALSE. I see that viaExaIsOffscreen() calls into libdrm, it never used to. Is that expected to work on setups where there is no DRM/DRI kernel stuff running? Thanks Daniel _______________________________________________ Openchrome-devel mailing list Openchrome-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/openchrome-devel
