Re: Cluster size

[Richard G. <rg...@ta...>]

lau...@fr... wrote:
>>>I describe again the other problem ( as you said, it can be kernel, but it
>>
>>is
>>
>>>really strange):
>>>1) Recording on a new track is always good (no xruns), even if there is lot
>>
>>of
>>
>>>tracks
>>>2) Re-recording is always bad (xruns), even on a few tracks
>>
>>I'm pretty sure this is due to wave scrolling/playback.  Can you try
>>re-recording with channels muted or with wave scrolling disabled?
>>
>>Richard.
>>
> Wave scrolling was disabled . With every tracks muted, still have xruns
> (only recording xruns in that case), may be less , need more tests.

I tried to reproduce it, but didn't succeed.  Though I have only limited
audio hardware, so recording/playing of more than two tracks is not 
possible for me.  Is there a minimum recording time before you see the 
problem?  Looking at profiles I don't see much differences between

  1. Recording into an empty stereo track (about a minute)
  2. Over-recording the whole stereo track with "Record into selection"
  3. Over-recording nearly the whole stereo track, but a little bit at
     the start and the end (again with "Record into selection")

For these tests I had playback and wavescrolling disabled, i.e. I was
just recording.  There weren't any XRUNs; I use alsa and a buffer size
of 192, though I think with playback disabled we use 32k as buffer size
if I remember correctly.

Maybe you can provide me with a recipie to reproduce the problem?

Thanks,
Richard.