Re: CVS commit: src/sys/dev

[Nathanial Sloss <nat%netbsd.org@localhost>]

Hi isaki@,

I'm currently making in kernel mixing optional depending on a sysctl variable 
to benefit slower/older platforms and as a default if audio auto config fails, 
it will work as audio did before in kernel mixing.

This commit was an intermediate works as I may have to inject audio into the 
mixring directly and as such it would be more usefull as a virtual channel.

If possible I will use the hardware ring directly in this case I will restore 
the record/play ringbuffers and reduce the memory consumption of this commit.

Best regards,

Nat

On Mon, 25 Sep 2017 22:58:56 Tetsuya Isaki wrote
> At Sun, 24 Sep 2017 23:40:41 +0000,
> 
> > Module Name:	src
> > Committed By:	nat
> > Date:		Sun Sep 24 23:40:41 UTC 2017
> > 
> > Modified Files:
> > 	src/sys/dev: audio.c audiovar.h
> > 
> > Log Message:
> > Move play/rec mix ring buffers into a virtual channel sc_mixring.  NFCI.
> > 
> > 
> > To generate a diff of this commit:
> > cvs rdiff -u -r1.406 -r1.407 src/sys/dev/audio.c
> > cvs rdiff -u -r1.64 -r1.65 src/sys/dev/audiovar.h
> 
> Is struct virtual_channel too large (sparse) to only have
> two ring buffers?  sizeof(struct virtual_channel) is 1244 bytes
> and one sizeof(struct audio_ringbuffer) is 92 bytes on recent i386.
> And I also think mixring is not (a part of) virtual channel.
> 
> However, I agree the name of sc->sc_pr/sc_rr is not easy to
> understand if you are focusing on the name.
> 
> Thanks,
> ---
> Tetsuya Isaki <isaki%pastel-flower.jp@localhost / isaki%NetBSD.org@localhost>