On Sun, 6 Mar 2005, Greg Stark wrote: > Mark Vojkovich <mvojkovi@xxxxxxxxxxx> writes: > > > That sounds like a driver issue. You might want to look through > > the cvs history for suspicious changes. This one looks suspicious: > > > > http://cvsweb.xfree86.org/cvsweb/xc/programs/Xserver/hw/xfree86/drivers/mga/mga_video.c.diff?r1=1.35&r2=1.36&sortby=date > > > > The description says: > > > > 85. Fix for Xv resolution loss problem with Matrox cards (Michael Lampe). > > > > but, if I recall correctly, some of the resolution loss issues > > with Matrox cards were because the video scaler didn't have enough > > bandwidth to display the full resolution. Reclaiming the resolution > > in bandwidth challenging situations probably results in fifo underflows > > which, at least on the NVIDIA cards that I'm familiar with, would > > produce results like you are describing. > > I don't follow that at all. What do you mean by "resolution loss" and > "reclaiming the resolution"? I don't remember the details because it was a long time ago, but back when I wrote the original MGA Xv support, I recall the driver using a lower-bandwidth mode where the chroma was subsampled in some way. I assume, but am not positive, that the mentioned patch is undoing that. > > Looking at that change it seems like it just removed a bunch of functionality > including the "accelerated 2x horizontal zoom" and didn't replace it at all. > It also changed a couple magic constants. That sounds like it removed the lower-bandwidth path. Perhaps you used to fall into the removed lower-bandwidth path? > > But I don't see how this code would make it work sometimes and not other > times. Specifically how it would work fine for a while and then stop working > completely. I'm assuming by not working, it looks like it's working from a client point of view, but the overlay just never shows. It's conceivable that overlay fifo underflow in the chip could cause the overlay to stop functioning. I've seen it happen on some NVIDIA cards. If this is happening, the OverclockMem option might help. > > It seems more like a higher level state corruption. Like maybe it still things > some other client is connected to the Xv port even though nothing is any more. > > Is there some way to query to see what X thinks is using the Xv scaler? Or > whatever other state is being kept? xvinfo doesn't seem to change output at > all when mplayer is running. Apps are supposed to grab the port when they use it. If another app tries to use it, they will fail to access the grabbed port. I believe mplayer grabs ports correctly. I'm running CVS from 2/22/05 and I don't notice any problems running MPlayer-1.0pre5. I have two NVIDIA cards in the system and I am using Xinerama. Mark. _______________________________________________ XFree86 mailing list XFree86@xxxxxxxxxxx http://XFree86.Org/mailman/listinfo/xfree86
