Subject: Re: [linux-audio-dev] Re: costs of IPC
From: Abramo Bagnara (abramo_AT_alsa-project.org)
Date: Tue May 15 2001 - 20:57:07 EEST
Abramo Bagnara wrote:
>
> Steve Harris wrote:
> >
> > On Tue, May 15, 2001 at 05:20:47PM +0200, Abramo Bagnara wrote:
> > > >
> > > > I'm a little puzzled. the "sp" case doesn't have any IPC costs or
> > > > context switches. How can it be consuming 12-15% of the CPU time in
> > > > these costs? These costs are always zero for this case, surely?
> > >
> > > As far as I understand Steve has not plotted the difference but the two
> > > CPU usage %. The relative cost of mp approach is the difference between
> > > the two values.
> >
> > Yes. When shared is 0, stack is constant, but not zero, and vica versa. So
> > osme work is being done.
> >
> > I think a 2d graph of shared against stack against cpu use would be
> > telling, but it would take a really long time.
>
> Stack is not a lot variable: we can easily make an high esteem in 4 or 8
> Kb.
>
> More interesting is shared and perhaps to add by component pollution:
> note that in sp each component have to pollute a *different* area.
>
> I don't believe this make some difference, but I'm too ignorant to be
> sure.
>
> I'll made some change and I'll try.
I've attached latest version (with I think reasonable default values).
-- Abramo Bagnara mailto:abramo_AT_alsa-project.orgOpera Unica Phone: +39.546.656023 Via Emilia Interna, 140 48014 Castel Bolognese (RA) - Italy
ALSA project http://www.alsa-project.org It sounds good!
This archive was generated by hypermail 2b28 : Tue May 15 2001 - 21:14:11 EEST