On Wednesday 19 May 2010 14:46:40 Valentin Eduardo (Nokia-D/Helsinki) wrote:
On Wed, May 19, 2010 at 01:36:51PM +0200, Ujfalusi Peter (Nokia-D/Tampere)
wrote:
On Wednesday 19 May 2010 13:52:27 Valentin Eduardo (Nokia-D/Helsinki) wrote:
The DMA pointer is easy, and it was know before as well, but according to my tests, the McBSP FIFO caused delay reporting is fairly accurate. I'll ask my users, if they have done some additional tests.
OK. fair enough. What if you really want to avoid the delay at all?
You can't. the McBSP FIFO can not be bypassed. The only thing one can do is to take it into consideration.
hmmm well, you can still do not use the fifo completely. Use only partially depending on latency requirement would still be possible.
Ah, you mean the idea, where in DMA completion interrupt we configure the McBSP threshold to maximum, and on McBSP FIFO empty interrupt, we configurre back the threshold to what the level we want to use from McBSP, and start the DMA? This all has to be done 'manually', and if there is _any_ latency in the kernel, than we have channel switch, glitch, pause, and whatnot.. I remember, we tried that, and well, did not worked that reliable.