Intel's MKL LaPack vs. ZTRMM: advice needed

Message

lahaye · #1 Post by **lahaye** » Tue Apr 25, 2006 3:18 am

Hello,

In VASP the GUIDE there is this information:
If the LAPACK is not well optimised, the call to ZTRMM should be avoided,
and replace by ZGEMM. This done by specifying NOZTRMM in the makefile.

I want to use the MPI version of VASP on a Linux/Intel Xeon cluster
(8 CPUs) with LaPack from Intel's Math Kernel Libary (MKL), e.g. in
the Makefile I have:

Code: Select all

LAPACK= -L/opt/intel/mkl/8.0.2/lib/32 -lmkl_lapack64 -lmkl

Is the MKL library "well optimised", or not? Should I use ZTRMM, or not?

Thank you,
R.

#2 Post by **admin** » Tue Apr 25, 2006 1:03 pm

in general it is. make a small benchmark test if you are not sure about your installation.