[Wien] machines file

Peter Blaha pblaha at zeus.theochem.tuwien.ac.at
Fri Apr 8 14:43:00 CEST 2005


> lapw1, but for executables lapw1_mpi !!  >Adding the -x switch on the first
> line of lapw1para gives you extra information).
> 
> We are trying to fix this. Could you explain more detail on this? Why can it
> pass lapw0_mpi but not lapw1_mpi? Thanks.

edit $WIENROOT/lapw1para and lapw0para
in the first line change the switch to   -xf
This will give you lots of debugging info and you need to gothrough this.

>  We did notice to what Prof. Blaha mentioned on the efficiency of mpi on dual
> node. The problem is not it will speed up the calculation or saves memory but
> when we run one wien2k job it will occupy all most > 99% CPU on one node. If
> in the same time someone run also wien2k, this will slow down the whole
> system. No one else can use host anymore :( because, in this case, 2 or 3
> nodes will be occupied with more than 99% CPU. Due to this we would like to
> run wien2k in parallel mode.

I'm still not sure I understand this:
if lapw1 occupies 99% of the cpu this should be very good.

Of course, if on a single node-cpu a second lapw1 us running, it will slow 
down to 50% (and if there is not enough memory, it will even crash).

On a dual cpu-node, a second lapw1 can run (provided you have enough 
memory). There will be a small performance loss (overloaded memory bus), 
but this is still the BEST you can get out of a dual node cpu.

Of course, while wien2k is running, nobody else should use these nodes (or 
performance/synchronization will go down. 

I do NOT see, how lapw1_mpi could help ??? It will only occupy more nodes 
with 100% (and acutally be slower) and even less other people can use the 
system.

Maybe you need a queuing system (pbs,...) or you may "renice" the wien 
jobs giving them less priority, as disscussed recently on this mailing 
list.



                                      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/
--------------------------------------------------------------------------




More information about the Wien mailing list