3 Sep
2013
3 Sep
'13
4:04 p.m.
On Thu, 29 Aug 2013, James Stone wrote:
At 32 frames/period (reported round-trip latency 1.33ms), it starts up but there are too many xruns for it to be usable.
After further testing (off-list), it turns out that 32 frames/period works okay with three periods/buffer instead of two. Not surprising really; with two periods you get only two URBs, and whenever one of them completes, the other isn't long enough to exceed the scheduling threshold.
In summary, it now appears that the most recent patch works better than the current driver.
Daniel and Eldad, more testing of the most recent proposed patch would be welcome.
Yes, please, more testing is needed. I intend to submit the patch when the current merge period ends.
Alan Stern