3 extra driver need to develop for GTA03

andrzej zaborowski balrogg at gmail.com
Wed May 28 16:55:54 CEST 2008


On 28/05/2008, Andy Green <andy at openmoko.com> wrote:
>  Somebody in the thread at some point said:
>
>  |>  > I like the idea of that userspace driver mux /dev/pts concept, but I
>  |>  > didn't like reading about weird autogenerated code.  Nobody said it
>  |>  > didn't work reliably though.  Does it actually work reliably or are
> we
>  |>  > supposed to mount an effort to do this in kernelspace?
>  |
>  | In the previous discussions reliability wasn't a concern with
>  | userspace muxer, the main concern was the performance and it's still a
>  | valid issue.  However, there was strong indication that someone was
>  | working on it for a different project (laforge + someone else for yet
>  | another project) and that it would come from upstream one day.  Since
>  | there's a working userspace implementation it stopped being urgent.
>
>  So we determined the performance is adequate for the gsm module we use,
>  or we suspect it isn't?  If we don't know, what's the plan for making sure?

I'm not authoritative to say what is the goal here for OM, I think
Mickey already said we determined the current implementation is fine
at this time.

In terms of bottlenecks I'm sure the muxer is not one, neither is the
gsm module, if you want to use the usual serivces (GPRS/EDGE).  I
think Harald did post some numbers from his experience in A780 project
where they actually have a working kernel-side TS0710 implementation
(written in a very A780 centric way unfortunately) including bit rates
at which the muxer potentially starts to be the bottleneck.  Below
that it's just some wasted cycles from going to userspace and back
more than necessary.

Cheers
-- 
Please do not print this email unless absolutely necessary. Spread
environmental awareness.




More information about the openmoko-kernel mailing list