[Rd] RE: MKL seems to beat ATLAS, but some problems... (was RE: [R] li nkin g R against MKL)

Prof Brian Ripley ripley@stats.ox.ac.uk
Thu, 17 Jan 2002 19:02:30 +0000 (GMT)

On Thu, 17 Jan 2002, Liaw, Andy wrote:

> Prof. Ripley,
> > From: Prof Brian Ripley [mailto:ripley@stats.ox.ac.uk]
> [snipped...]
> > R cannot currently make use of anyone else's LAPACK routines (because
> > we've found too many problems with them).
> >
> > If you are getting no useful diagnostics, try
> > dyn.load("RHOME/modules/lapack.so").  That ought to tell you
> > why you can't
> > load it.  My guess would be to do with libraries, so do an ldd on
> > lapack.so and check that all the dependencies are resolved as
> > you expect.
> I wonder if the problem is the clash between the Lapack routines in MKL and
> those packaged with R.

> Sorry, but seems like I need a bit more spoon-feeding:
> I did the dyn.load, and got:
>         unable to load shared library
> "/home/andy/R-patched-mkl/modules/lapack.so":
>   /home/andy/R-patched-mkl/modules/lapack.so: undefined symbol: lsame_

Right, that's it.  LSAME is a BLAS routine, and your BLAS doesn't have
it.  There is a version in src/appl/blas.f, but you don't seem to
have a complete BLAS to link against.

If you really have a complete LAPACK3 (which I would not bet on)
you can try building in src/modules/lapack omitting double.f and
complex.f: you will need to rename a few entry points in Lapack.h
as they have already been changed to avoid conflicts with sunperf
(I think).  See also the comments in the README there.


> Looks like all the libraries it needs are there, but why can't it find those
> undefined symbols?

They are in R.bin itself.

Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch