At Tue, 8 Sep 2009 17:34:35 +0200,Giovanni Maruzzelli wrote:> > On Tue, Sep 8, 2009 at 5:10 PM, Takashi Iwai<tiwai@xxxxxxx> wrote:> > Don't pass '=', e.g.> >> > # echo channels_max 3 > /proc/..../dummy_pcm> >> > Thanks, with your hint I was able to write to the proc file.> > I confirm that the Skype client works well if I put (writing to the> proc file, without modifying the driver source): buffer_bytes_max=1024> and period_bytes_max=512, and that it works bad if I do not write on> the /proc file.> > It would be very very very nice if you add this proc interface (or> another method to alter those buffer and period values) also to the> non-debug build, so users can be able to use the standard distributed> prepackaged ALSA drivers, without resorting to editing source or> recompiling. Well, I have some uneasy feeling to expose it as non-debug.With proc parameter, you can pass every crazy value that can screw upthe PCM core. In the non-debug mode, some parameter checks areomitted by assumption that the driver gives all correct, thus it mightgive Oops at worst. Although its write access is limited to rootonly, a possible cause should be removed as much as possible. > BTW: Would you mind to bump the MAX_PCM_SUBSTREAMS to 128? Currently> is 16, and is a cap to the substreams actually created, by a> configurable param default to 8. (people use routinely 20 concurrent> Skype clients, 40 substreams. But there is juice in modern servers to> have some more. 64 seems roomy enough [Skype client itself uses lot of> cpu]). That's easy and should be safe. I'll change later. thanks, Takashi_______________________________________________Alsa-devel mailing listAlsa-devel@xxxxxxxxxxxxxxxxxxxx://mailman.alsa-project.org/mailman/listinfo/alsa-devel
