--- In liblf-dev@yahoogroups.com, Bjorn Roche <bjorn@...> wrote:
>
>
> On Oct 30, 2008, at 5:10 PM, James McCartney wrote:
>
> > On Wed, Oct 29, 2008 at 11:37 AM, Bjorn Roche <bjorn@...>
> > wrote:
> > > Plus, gcc, for example, tends to
> > > insert memory barriers in places, like at function calls.
> >
> > I really don't think so. citation?
> >
>
> I can't find a citation right now, so I could be wrong. If I am wrong,
> sorry about that. It's obviously not behavior to count on anway, since
> even if you know you are always working with gcc functions can be
> inlined.

I've read so many citations, and other theories about ring buffers and
memory barriers. But to me they're useless until someone can write a
piece of code that fails when barriers are absent and succeeds when
they're in, and thus proves they are needed.

Theories should help writing such a test case, but for now, I (and
others on linux-audio-dev and jack-devel) couldn't figure that out. Or
better said, we thought we did, but running it on the most vulnerable
hardware we could think about (PPC SMP), we observed no failure.

Any link, suggestion?

--
Olivier Guilyardi / Samalyse