> > I was assuming that using poll (snd_pcm_wait), with the > > alsa-lib/test/latency.c program would have snd_pcm_read > > return period size samples at a time, but this is not > > the case. > > [..SNIP..]
> It's intended. It looks better in my eyes, if we can process some data > ahead in low-latency application that wait for whole batch. Of course, > optimized applications for cpu-usage might work differently. I see... However, I can imagine many situations in which you are processing with a fixed buffer size (doing an fft for example). And I still think that it is reasonable to expect that the default behavious is that readbuf will always return the same amount of samples. Would it not be possible to add a commandline argument to specify the behaviour ? -p,--poll use poll (wait for event - reduces CPU usage) events are generated when a new period of frames is available -q,--quickpoll use poll (wait for event - reduces CPU usage) events are generated when new frames are available or something like that. As the user has to specify either one, he/she will be concient of what is happening. Apart from that, I think it is a bit agressive that the default execution of the latency test is non-blocking, which causes all CPU to go to the latency test, and at first sight really looks like your machine locks up. Would it be a good idea to make blocking the default, and change the argument -b,--block block mode for -n,--nonblock non-block mode ? Maarten _______________________________________________ Alsa-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/alsa-devel