[arch-general] Opinions on pulseaudio [WAS: PulseAudio in [testing]]

fons at kokkinizita.net fons at kokkinizita.net
Mon Nov 29 14:14:28 CET 2010

On Mon, Nov 29, 2010 at 10:31:06AM +0800, Ng Oon-Ee wrote:

> Thanks fons, I was under the mmap alsa capability implied direct control
> over the hardware's memory.

It probably does in some cases (e.g. a hw: device on a PCI bus),
but only to the sample data - there's no direct access to any
control registers ever in ALSA. 

Since the access _could_ be to real HW  it's of course ALSA
that dictates the layout (separate buffers for each channel, or
interleaved, or even more complex schemes) and the sample format
(16/24/32 bit or float, LE or BE). The client application just
has to accept this and do any conversions itself. 

Maybe the PA devs were thinking they had to emulate all of those,
which could be a nightmare but is not necessary. They could just
select the format that corresponds to any buffers they already
have, and only support that one. Clients using mmap mode check
what is available and adapt to it. 



There are three of them, and Alleline.

More information about the arch-general mailing list