[daip] Parallel AIPS?
Eric Greisen
egreisen at nrao.edu
Wed Apr 22 10:38:34 EDT 2009
Karl Menten wrote:
> Eric,
>
> Attached is an email announcing the opportunity to competitively get
> time on super computers. I though about applying just to get acquainted
> with "the system". However, this would really only make sense if some
> sort of parallelized AIPS existed. Is anything useful available?
>
At present we are still anticipating rather than doing parallelization.
Bill Cotton's obit package has done some parallel tests and we expect to
do some later. But Bill's tests on multi-threading suggest that it has
to be explicitly coded rather than automatic threading and the cluster
computing also has to be explicitly coded. I am considering designs in
which a user can spread his problem to multiple cpus - but the user will
have to control it I expect. All this may not be of interest on a super
computer.
Eric Greisen
More information about the Daip
mailing list