On Thursday 04 Jul 2002 7:11 pm, Keith Whitwell scribed numinously:" > Tim Smith wrote: > > On Thursday 04 Jul 2002 5:28 pm, Keith Whitwell scribed numinously:" > > > >>>I've just tried two glxgears at the same time and I am seeing utterly > >>>strange behaviour with two glxgears running, where sometimes if I move > >>>one it will leave bits of gears behind (sometimes a whole frame), and > >>>sometimes corrupt my desktop background image (and I mean really > >>>corrupt; not a redraw error) with bits of white, indicating a write > >>>into an off screen area; the bitmap is tiled and if I refresh the > >>> whole desktop the corruption shows up in all of them; I have to > >>> upload a new background image to make it go away. > >> > >>This is the cliprect issue... > > > > Arrgh. This happened to me *once* only. I haven't been able to make it > > happen again since. (FWIW the background corruption occasionally > > happened before I started playing with the tree) > > > > Using the SET_SCISSORS command, I am getting what I'd expect (now that > > you've explained the coarse scheduling): I once managed to move the > > glxgears window such that the gears are being drawn off-center (i.e. > > where they used to be before the window moved) but they were clipped > > correctly at the edge of the window. > > > > With two glxgears overlapping, I get a clean, properly clipped display > > when the rear one is animating. I see a slight flicker in the overlap > > region when the front one gets the lock, but otherwise everything looks > > fine. I've tried rebooting and power-cycling too. Most annoying. > > The slight flicker is all I usually see either. This is why I didn't > spot it in cursory testing. However, if you disable > radeon_emit_clip_rect() altogether, ie put a 'return' in the first line, > you get exactly the same behaviour... The coarse sheduling plus the fact > that cliprects keep on working normally for 'clear' and 'swapbuffers' > manages to hide the fact that they aren't working at all for 3d > rendering.
Yes. It looks like the SET_SCISSORS command doesn't do what I think it does. Back to the drawing board :-/ There may yet be a clean way to fix my other lockup without having to do wait states. At least I have another test I can use to make sure I haven't broken anything this time. -- Tim Smith ([EMAIL PROTECTED]) "Violence never solved anything permanently." -- Genghis Kahn ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Caffeinated soap. No kidding. http://thinkgeek.com/sf _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel