[R-sig-hpc] multithreaded blas. can you smarten me up?

[Dirk Eddelbuettel]

On 2 March 2010 at 11:26, Paul Johnson wrote:
| I have been searching for advice about installing R on quad core
| systems that exist within a cluster and I came to a post in r-help.
| The thread "[R] (performance) time in Windows vs Linux" was pretty
| long, and at one point Professor Ripley observed:
| 
| On Mon, Jun 29, 2009 at 1:59 AM, Prof Brian Ripley
| <ripley at stats.ox.ac.uk> wrote:
| 
| >
| > The only way that a single R process will be using more than one CPU is if
| > you have added a mulithreaded BLAS (and I've never heard of one being used
| > successfully with R for Windows) or other add-on such as Luke Tierney's
| > pmath[0] packages.  Packages such as snow and multicore run multiple R
| > processes.
| >
| > I do run a multithreaded BLAS on my 8-core Linux box and do often see 'top'
| > well over 100% -- I just tested and saw 798.9%.
| >
| 
| Clearly, I need some multithreaded BLAS library. After installing
| that, one must re-compile R with the configure option to force it to
        ^^^^^^^^^^^^^^^^^^^^^

Not so. "The nice thing about standards...." as the quip goes. 

Blas is one.  So you take your _existing_ R installation, swap the blas and
you're done.  
 
That is one reason why we _like_ dynamic linking. You can substiture
libraries at well (provided the API/ABI stays the same).

| Does anybody know if it is then necessary to force a re-compile of all
| R packages as well?  (here's me wishing they would just take advantage
| of R's access to that BLAS, but I doubt it).

Likewise, you don't.  There is some good advide in the R Admin manual.

Dirk

-- 
  Registration is open for the 2nd International conference R / Finance 2010
  See http://www.RinFinance.com for details, and see you in Chicago in April!