On Thu, 2006-06-01 at 16:38 +0900, Matthieu Redouin wrote: > On Thu, 2006-06-01 at 08:52 +0200, Clemens Ladisch wrote: > > Matthieu Redouin wrote: > > > I am trying to have short latency on my computer. I am using the > > > hda-intel driver and I am wondering about the output latency : > > > - with a buffer size of 64 and period of 2, I get around 1.5ms of output > > > latency > > > - with a buffer size of 64 and period of 4, I get around 4.2ms of output > > > latency ! I don't figure out this big difference. > > > > What are the units you're using here? Is your period size in bytes, in > > frames, or the number of period per buffer? And what are the sample > > format and rate? > sorry, I am using frames for the buffer size and for the period, it is > the number of period per buffer. > The sample format is 32bit LE and the rate is 48000Hz (and 2 channels). > > > > Setting period size to 4 allows me to set buffer size to 16 and have an > > > output latency of 1.2ms, whereas period size of 2 and buffer size of 16 > > > gives me too many xruns (even with patched kernel). > > > So I am wondering if buff/period size of 64/2 was my hardware limits and > > > 16/4 settings only a trick handled by the driver ? > > > > > Have a look into /proc/asound/card0/pcm0p/sub0/hw_params to see what the > > parameters used by the driver are. > thanks !!! > It helped a lot. > So in fact, my problem is not really about buffer size, but xruns : > - when I start Jack with -p16 (period size in frames) -n4 (number of > period), I get no xruns > - when I start with -p32 -n2, I have a lot ! and when I start with -p16 -n2 (which is the limit of my hardware if I read correctly the driver), I don't have any xruns but jackd is killed by the watchdog after a few seconds.
> why ? shouldn't it be the same thing ? And my kernel is already patched. > > thanks for the tips, > matt _______________________________________________ Alsa-user mailing list Alsa-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/alsa-user
