[Rcpp-devel] Install RcppArmadillo against Intel MKL

[Zarrar Shehzad]

Great. This definitely seems like a better approach to rely on R. I have
Revolution R setup on another machine so I just installed RcppArmadillo with
that and indeed it appears to now be relying on Intel's MKL.

Thanks!
Zarrar

On Sat, Sep 3, 2011 at 12:45 AM, Dirk Eddelbuettel <edd at debian.org> wrote:

>
> Hi Zarrar,
>
> On 2 September 2011 at 17:19, Zarrar Shehzad wrote:
> | Is it possible to install RcppArmadillo with Intel's MKL library?
>
> In principle, yes. In practice, probably too as well :)
>
> When we started this we used Armadillo as library and via a configure
> step. Doug later convinced us that relying on R is easier -- so we can take
> LAPACK and BLAS and other things as given. This then, we pre-configure
> Armadillo and essentially just use templates, which makes portability a
> snap.
>
> So you should "just" need to declare MKL the proper way for R itself, and
> as
> RcppArmadillo calls into R to get BLAS resolved, you should get MKL that
> way.
> As MKL is multithreaded so you should see a speed pickup for certain ops.
>
> See the R Inst + Admin manual for how to setup R and and MKL.
>
> | I tried to replace the "$(LAPACK_LIBS) $(BLAS_LIBS)" in the Makevars with
> links
> | to Intel's MKL: "-L/opt/intel/composerxe/mkl/lib/intel64/ -lmkl_gf_ilp64
> | -lmkl_core -lmkl_def -lmkl_gnu_thread -
> | lgomp -L/opt/intel/composerxe/mkl/lib/intel64/ -lmkl_gf_ilp64 -lmkl_core
> | -lmkl_def -lmkl_gnu_thread -lgomp" and also uncommented the "#define
> | ARMA_USE_LAPACK" and "#define ARMA_USE_BLAS" in
> armadillo_bits/config.hpp. Are
> | these the wrong things to do or is there something else I should be
> doing?
>
> Try leaving it as is.  At least on my Debian / Ubuntu systems, it would
> transparently pass through to BLAS as used by R.
>
> Dirk
>
> | Thanks
> | Zarrar
> |
> | ----------------------------------------------------------------------
> | _______________________________________________
> | Rcpp-devel mailing list
> | Rcpp-devel at lists.r-forge.r-project.org
> | https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
>
> --
> Two new Rcpp master classes for R and C++ integration scheduled for
> New York (Sep 24) and San Francisco (Oct 8), more details are at
>
> http://dirk.eddelbuettel.com/blog/2011/08/04#rcpp_classes_2011-09_and_2011-10
>
> http://www.revolutionanalytics.com/products/training/public/rcpp-master-class.php
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20110903/fdfefa27/attachment-0001.htm>