[Wien] Compiling lapw1_mpi with HP-mpi and MKL

[Peter Blaha]

I just want to summarize the solution to this query which we reached after
some trial and error.

According to a summary by B.Monnet the main change was a change of a Kernel
parameter in their Linux-Server. (This could provide a hint to similar problems
at other big cluster installations).

Finally they ran WIEN on 1000 cores....


--------------------------------------------------
PS: Regarding the configuration used to have wien2k successfully running
on the cluster (128 HP Proliant dl160 with 2 Xeon E5472 @ 3.00GHz & 64
GB of RAM, IB DDR):

I've made few modifications on the cluster. I found that kernel.shmax
was too small to run large mpi jobs (kernel.shmax is now 8000000000)

I've modified lapw0 to replace mpi_address with mpi_get_address and also
change the param size INTEGER -> INTEGER*8
I've fully recompiled the application using Intel Fortran 11.0.074 with
-mp1 -prec-div -prec-sqrt to maximize the precision during floating
point conversion.
I've added MKL_SERIAL=Yes & OMP_NUM_THREAD=1 during the run.

Lombardi, Enrico schrieb:

> We are trying to compile mpi-parallel Wien2k lapw1/2 on an infiniband 
> system, but have not been successful up to now.
> 
> We would appreciate an indication of which combinations of mpi-library, 
> math-library and compiler are known to work on infini-band systems?  
> Also what scaling has been achieved on such systems up to now?
> 
> Currently, we are compiling using different scenarios:
> 
> 1. HP-MPI v2.3.1, Intel Fortran v 11.0 and MKL :  In this case the code 
> compiles without error messages, but lapw1 crashes immediately with 
> numerous segfaults.

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------