From: Michael R. <ama...@fr...> - 2002-03-30 14:33:03
Attachments:
dxr3.patch
deny_backward_adjust.patch
|
I did some work on the dxr3 plugins again. Enclosed is a patch against current cvs which fixes the following issues: * make dxr3 work again by updating to current video_out interface * fix some problems when switching from non-mpeg content (e.g. the xine logo) to mpeg content * handle situations when video_out tells the dxr3 decoder to skip frames The last point is the most problematic one. Basically, when playing mpeg content, the dxr3 decoder plugin is unable to skip frames, since we are feeding the card with a consecutive mpeg stream of which we can't take single frames out. So my soultion here is to adjust vpts_offset of metronom via the lately introduced metronom option so that video_out doesn't want us to skip frames any more. The problem with this solution is that when the dxr3 decoder adjusts vpts_offset to its own needs, the audio_out would turn it back. Therefore I had to introduce a new metronom option which, when set, only allows positive adjustments to vpts_offset. The corresponding patch to metronom is attached (warning: the dxr3 patch results in compilation errors without it). I would appreciate comments on my work. If you think, my solution is clean enough I would even like to see it included into cvs before the release of 0.9.9. I have done some testing with it and for me, the dxr3 works better than ever before. All syncing problems are gone. Michael Roitzsch |
From: Rich W. <rj...@he...> - 2002-03-31 15:02:25
|
On Sat, 2002-03-30 at 14:27, Michael Roitzsch wrote: > I did some work on the dxr3 plugins again. Enclosed is a patch against > current cvs which fixes the following issues: > > * make dxr3 work again by updating to current video_out interface > * fix some problems when switching from non-mpeg content (e.g. the xine > logo) to mpeg content > * handle situations when video_out tells the dxr3 decoder to skip frames Great work! I have comitted some of the fixes (the ones that make the Dxr3 work again) but have left the metronom changes to be looked at by someone more familiar with it. |
From: Michael R. <ama...@fr...> - 2002-03-31 16:53:16
Attachments:
dxr3.patch
|
I had not expected that my patches would make it into cvs that fast. (My last one took me three postings.) So at first, a big thank you for that. There is just one little problem I just discovered: I have just tried my dxr3 changes without the metronom patch (because I do not consider it very clean) and it still works. Strange enough, it works even better. I do not really know why, but it seems, that the audio drift patch commited earlier by Miguel Freitas helps a lot with dxr3 sync. I have to do some more testing, but I really think, that the new metronom option introduced by my patch is not needed. I am sorry for the inconveniences caused by me but I think we should revert the patches I sent in before. I have prepared a new one which you can apply after reverting. It includes all the fixes of my previous one (plus one more) but it doesn't use the metronom option. If you want, I can also provide a patch against current cvs doing all this in one, without the reverting. I will do some testing this evening, if sync is really fine without patching metronom. Sorry again. Michael Roitzsch |
From: Miguel F. <mi...@ce...> - 2002-03-31 23:57:41
|
Hi Michael, Michael Roitzsch wrote: > There is just one little problem I just discovered: I have just tried > my dxr3 changes without the metronom patch (because I do not consider > it very clean) and it still works. Strange enough, it works even > better. I do not really know why, but it seems, that the audio drift > patch commited earlier by Miguel Freitas helps a lot with dxr3 sync. Well, i must say i'm very happy to hear that! :) Both because the success story with audio drift patch and the not needed metronom "hack"! great! > I have to do some more testing, but I really think, that the new > metronom option introduced by my patch is not needed. > I am sorry for the inconveniences caused by me but I think we should > revert the patches I sent in before. I have prepared a new one which > you can apply after reverting. It includes all the fixes of my previous > one (plus one more) but it doesn't use the metronom option. If you > want, I can also provide a patch against current cvs doing all this in > one, without the reverting. Please, if you can provide a complete patch against cvs i would apply it by tomorrow. (i will have some time to do xine stuff, including libmpeg2 updates) regards, Miguel |
From: Michael R. <ama...@fr...> - 2002-04-01 12:02:23
Attachments:
dxr3.patch
|
Hi Miguel, Miguel Freitas wrote: > Michael Roitzsch wrote: > > There is just one little problem I just discovered: I have just > > tried my dxr3 changes without the metronom patch (because I do not > > consider it very clean) and it still works. Strange enough, it > > works even better. I do not really know why, but it seems, that the > > audio drift patch commited earlier by Miguel Freitas helps a lot > > with dxr3 sync. > > Well, i must say i'm very happy to hear that! :) > Both because the success story with audio drift patch and the not > needed metronom "hack"! great! Your audio drift patch is real magic. I did some extensive testing last night without my metronom option and xine with dxr3 performed amazingly great. I watched two dvds, for one of them the complete movie and full bonus material. Xine went up to a scr value of 1.5 billion without the need to restart it and, best of all, without the slightest sync problem. It recovers from seeking and even from fast forward. The audio drift code adjusts vpts_offset so smooth, that it won't be noticable and so fast, that even menu transitions are in sync. Great stuff! Xine really outruns the windows player shipped with my dxr3 in terms of sync now. > Please, if you can provide a complete patch against cvs i would apply > it by tomorrow. (i will have some time to do xine stuff, including > libmpeg2 updates) The patch attached removes the new metronom option and fixes a strange lockup in the em8300 driver. I would say, from the dxr3 point of view, 0.9.9 is ready for release. Michael Roitzsch |