[ardour-dev] Ancient history?

[cv223 at comcast.net]

Hi,

I've been using/learning Ardour for a little while for recording with a Delta 1010LT in a SMP system using a SCSI Raid0 drive.  I've been having xrun troubles while recording with the setup, which I've posted about on the linux-audio-users list (http://music.columbia.edu/pipermail/linux-audio-user/2004-October/016676.html).  I won't reiterate the whole problem here, just point out that I can record with no xruns to an IDE drive with the same jack settings (-n 2 -p 256 -r 44100) that generate dozens of xruns per minute on the Raid0.

I then ran some tests using ecasound, which seemed to indicate that the problem was not _simply_ the SCSI/Raid0 (http://music.columbia.edu/pipermail/linux-audio-user/2004-October/016812.html).  I can record all 8 inputs and write to either drive with no xruns, using a buffer size of 128 and a sampling rate of 48000.  I can even write twice as much data to the harddrive (i.e., writing 2 copies of the audio) during recording without problems.

In my Ardour work through all this, I had been using versions up to 0.9beta11.2.  I just built 0.9beta19 and the xruns on the Raid0 virtually went away! Not entirely, but now I only see a couple per 1 minute recording.  I did a sort of 'binary search' through the intervening versions and found that the largest transition occured between 0.9beta11.2 and 0.9beta12 (a handful of xruns per min with the latter), with some additional improvement through 0.9beta19.  In addition, I had numerous xruns on both drives when running 2.6.x kernels (lately, 2.6.9-rc2-mm4-VP-S7), which also greatly decreased on both drives with the 0.9beta19 version.

I would really like to find out what changed between these two earlier versions and thought I would ask before wading through the change logs and the code (I've done a good bit of C and C++ programming, although not with audio).  This is partly out of curiousity (I can simply record to the IDE drive), partly out of the desire to see if recording to the Raid0 can be as good (or better!) as to the IDE, and partly to contribute to the community with some testing, if it is at all valuable.

Any insight would be greatly appreciated.  Perhaps this is a known issue (I haven't seen anything in the list archives, though haven't searched exhaustively) - if so, my apologies.  I've really enjoyed using Ardour so far, and it just seems to keep getting better!

Thanks,

Joel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ardour.org/pipermail/ardour-dev-ardour.org/attachments/20041101/3cb7c404/attachment-0002.htm>