El Domingo, 3 de Julio de 2005 10:34, Martin Wache escribi?: > Udo Richter schrieb: > > Luca Olivetti wrote: > >>I also wonder how is it > >>possible that vdr hasn't enough data to give to the device: no matter > >>how big the buffer, once it is filled it should stay more or less full, > >>unless you're at the end of the file. > > > > One guess: On playback start, buffers are empty and can accept lots of > > data. This could result in some busy loops until enough data is read to > > fill the buffer. > > > > If, for some reason, the disk read requires much CPU time, then the busy > > loop will actually slow down the disk reading too. > > > > Maybe, in case of softdevice, this is oscillating: A busy file reader > > takes CPU from decoding, causing decoding to stall. And as soon as the > > decoder is catching up, maybe by dropping frames, the file buffers run > > empty causing the file reader to busy loop again. (speculative, though) > > I guess this is a possible scenario, at least it explains the symptoms > quite acuratly. > Hmm, to finally solve this I guess we have to wait until Klaus answers. > I think he is now on his holidays, right? So I'll wait until he is back. > But thanks for your answers, at least I know now that the softdevice is > not the only one which has problems like this. > The patch also work well with vdr-xine with a athlon 64 3200. Without the patch there is a heavy cpu load when playing recordings. Before I use the atached patch with have similar effects. Jose Alberto -------------- next part -------------- A non-text attachment was scrubbed... Name: vdr-10_livelock.dpatch Type: application/x-shellscript Size: 674 bytes Desc: not available Url : http://www.linuxtv.org/pipermail/vdr/attachments/20050722/e7632b80/vdr-10_livelock.bin
