I'm using a multithreaded BLAS library with R and I see the expected speed improvements with matrix multiplication, svd, etc. However, glmnet continues to use only a single CPU. Since this package is compiled from Fortran, is this the expected behavior or is there a way to compile the glmnet package so that it uses the multithreaded BLAS library? Thank you Brian
BLAS with glmnet
2 messages · Brian Pasley, Peter Dalgaard
1 day later
On Apr 12, 2013, at 22:49 , Brian Pasley wrote:
I'm using a multithreaded BLAS library with R and I see the expected speed improvements with matrix multiplication, svd, etc. However, glmnet continues to use only a single CPU. Since this package is compiled from Fortran, is this the expected behavior or is there a way to compile the glmnet package so that it uses the multithreaded BLAS library?
You'd need to check with the author for full details, or study the code yourself. Or the JSS paper. As a general matter, a package won't benefit from a fast BLAS if it doesn't use it. Presumably, glmnet does not, or does not do so in timing-critical areas. I wouldn't expect Fortran per se to have anything to do with it, although I can't rule out that the code might be rewritten to take advantage of the BLAS. -pd
Peter Dalgaard, Professor, Center for Statistics, Copenhagen Business School Solbjerg Plads 3, 2000 Frederiksberg, Denmark Phone: (+45)38153501 Email: pd.mes at cbs.dk Priv: PDalgd at gmail.com