[R-sig-Debian] Compiling R-2.11.0 with ATLAS-tuned BLAS and LAPACK

[Andrew Piskorski]

On Fri, Jun 04, 2010 at 01:45:34PM -0500, Paul Johnson wrote:

> mm <- matrix(rnorm(10^7), ncol = 10^3)

> ### Following DOES cause use of more cores
> system.time(mmm <- mm %*% t(mm))

> 1. Will an R call to BLAS automatically spread out across cores?

Didn't you just answer that question above?  From your result, it
seems pretty clear the answer is that any BLAS will internally uses
multiple threads is determined entirely by that BLAS library's own
code.  What you use to call call that BLAS library, whether R or
anything else, is not relevant.

> It describes optimization with parallel BLACS and "scalapack".  If I
> want to see that in action, though, I have to purchase the actual
> Intel product, right?

BLACS and Scalapack are both open source, which googling will quickly
tell you.  Scalapack and the MPI libaries you need are also packaged
for Ubuntu, so just install them on your cluster and run.  Intel's MKL
is of course a commercial product.

The PLAPACK guys say that it is better than Scalapack, but it's not
pre-packaged by any Linux distros that I know of.  You'd probably need
to build it yourself.

-- 
Andrew Piskorski <atp at piskorski.com>
http://www.piskorski.com/