[R-sig-hpc] Problems parallelizing glmnet
Brian G. Peterson
brian at braverock.com
Thu Sep 6 19:45:24 CEST 2012
On 09/06/2012 12:20 PM, Zachary Mayer wrote:
> I was not aware of this unfortunate limitation. Was it specific to a
> particular back-end, or just foreach in general?
>
> On Thu, Sep 6, 2012 at 1:15 PM, Patrik Waldmann
> <patrik.waldmann at boku.ac.at>wrote:
>
>> >I would like to avoid foreach since we showed earlier that it is VERY slow.
Foreach is 'very slow' (i.e. has a lot of overhead) if your function
evaluation to be parallelized in computationally trivial, in which case
the benefits of parallelization are rather questionable.
In the case where your function evaluation is not trivial, the overhead
of foreach is marginal in comparison to the various *apply* methods that
it wraps, and almost certainly worth the flexibility to not change code
as your parallel backend changes.
--
Brian G. Peterson
http://braverock.com/brian/
Ph: 773-459-4973
IM: bgpbraverock
More information about the R-sig-hpc
mailing list