Ronan Keryell <Ronan.Keryell@xxxxxxxxxx> writes:
>>>>>> On Mon, 18 Feb 2013 18:46:05 +0100, Thomas Rast <trast@xxxxxxxxxxxxxxx> said:
>
> Thomas> The actual programming must be done in C using pthreads
> Thomas> for obvious reasons.
>
> Are there obvious reasons OpenMP would not be enough to do the job?
>
> It looks like a trade-off between the code readability & portability
> versus the real expressiveness of what parallelism control details are
> needed.
Except for the added dependency you mean?
I'm not sure exactly what the capabilities of OpenMP are that would help
here, but most likely it would work. It wouldn't really change the
amount of work needed, though, since the main work is in shuffling
around the existing code paths to be amenable to parallel access in the
first place. A "dumb" parallelization (i.e., just locking around all
shared structures) POC yielded very little speedup because of lock
contention.
--
Thomas Rast
trast@{inf,student}.ethz.ch
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html