compiling R | x86_64 | CentOS 5.3 | lapack/blas woes

Posted by cooch17 on 2009/6/26 4:36:40

Probably overly narrow question to generate much 'traffic', but here goes...

I recently upgraded my multi-Opteron machine to CentOS 5.3 (from FC 8). Went fine. A major app for me on this box is R (statistical software). Normally, I do a compile from scratch, using standard lapack and ACML BLAS. However, now matter how much fooling I've done, whenever I try to do the compilation, and run the make check, it throws a major error on some high-level matrix functions. The error method consistently points to lapack as the culprit. If I take lapack out of the compilation, then make check runs fun. Good right? Well, without lapack, several key programs take forever to run.

Anyone have any thoughts/comments/suggestions?

Thanks!