Ryan C. Thompson
rct at thompsonclan.org
Sat Nov 17 21:56:27 CET 2012
In reply to:
On 11/16/2012 09:45 PM, Steve Lianoglou wrote:
> But then you have the situation of multi-machines w/ multiple cores --
> is this (2) or (3) here? How do you explicitly write code for that w/
> foreach mojo? I guess the answer to that is that you let your "grid
> engine" (or whatever your SGE equivalent is that you might be running)
> handle that for you.
You just use makeCluster to spawn multiple workers per host.
host.cores <- c(host1=4, host2=8, host3=2)
cl <- makeCluster(rep(names(host.cores), host.cores))
Honestly, when I think about multi-machine parallelism, I generally
assume that multiple cores per machine is implied. In any case, the
point is that foreach supports any parallelization situation (multicore,
multi-machine, both) for which a parallel backend exists. So of all the
parallelism stuff that exists in R today, I think it's the closest thing
we have to a general parallelization framework.
More information about the Bioc-devel