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

[Paul Johnson]

On Tue, Jun 1, 2010 at 10:56 PM, Dirk Eddelbuettel <edd at debian.org> wrote:
>

> |
> | Here's what I don't understand about the current R packaging.
> |
> | In the deb packages,  I download from CRAN, there is no file
> | libRblas.so, so I don't see how to make that new link to test  a
>
> Who says you need libRblas.so?  We no longer do.
>
> | different BLAS.
> |


The R install & admin manual says so, actually.  I think that you know
what's going on much more accurately than it does, and perhaps you
don't see that doc the way we do.

Please look at this part to see what I mean

A.3.1.5 Shared BLAS

"Note that the BLAS library will be used for many of the add-on
packages as well as for R itself. This means that it is better to use
a shared/dynamic BLAS library, as most of a static library will be
compiled into the R executable and each BLAS-using package."

The R I&A document says that unless you have libRblas.so, then BLAS
support is statically compiled into R. In that case, a new BLAS would
require all R to be recompiled and  all the packages have to be
recompiled.

I suppose from what you say that those docs are out of date.  But I
still don't understand why the test described in README.atlas  works.

In the example Atlas readme you refer to, the administrator is assumed
to actually remove/purge whole packages and reinstall to test the
effect of these changes.  R "somehow" knows to use the new BLAS.  I
guess that means we are either relying on
1. Some post-installation script from the atlas debian packaging that
manipulates the library paths, or
2. R picks a BLAS at runtime.

Unless there is libRblas.so, how could I specify the symbolic link
that tells R which blas dynamic library to use?  (Without removing all
but one of the BLAS libraries that are installed).  On our production
system, there might be 4 or 5 editions/types of BLAS installed.

Note that README.atlas doc does recommend to the user that Atlas
should be recompiled to suit the particular system, explaining the
OP's curiosity about re-building Atlas.

pj
-- 
Paul E. Johnson
Professor, Political Science
1541 Lilac Lane, Room 504
University of Kansas