[DFTB-Plus-User] DFTB+ maximum number of threads
benjamin.hourahine at strath.ac.uk
Tue Mar 15 08:56:43 CET 2016
As Bálint says, MKL is probably the limiting factor. Are these 12 cores
or 6 physical cores with hyperthreading (i7 processor)? MKL uses the
physical core count, as there is no advantage for dense operations in
threads sharing a core:
If you really want to enforce a thread count of 12 you might also want
to try the MPI parallel version.
On 15/03/16 07:14, Bálint Aradi wrote:
> Dear Anton,
>> used by program can be successfully controlled by OMP_NUM_THREADS
>> environment variable from 1 to 6, so OpenMP seems to work fine. However,
>> if I try to use more (ideally I would like to use all 12 threads to
>> speed up calculations, I'm trying to work on pretty large cell), it
>> never does. So, the question is, if there are hard coded limits on the
>> maximum number of threads the code can use, or there is some trick to
>> compilation process to make it use more than 6 threads?
> We do not impose any hard coded limit on the number of threads.
> Actually, most threaded activity happens within MKL itself, so it is
> MKL, which apparently refuses to use more threads. But the question is
> really, does your machine recognise 6 or 12 threads? Usually MKL
> launches as many threads, as the OS reports cores on the machine.
> Best regards,
> DFTB-Plus-User mailing list
> DFTB-Plus-User at mailman.zfn.uni-bremen.de
Dr. B. Hourahine, SUPA, Department of Physics,
University of Strathclyde, John Anderson Building,
107 Rottenrow, Glasgow G4 0NG, UK.
+44 141 548 2325, benjamin.hourahine at strath.ac.uk
2013/14 THE Awards Entrepreneurial University of the Year
2012/13 THE Awards UK University of the Year
The University of Strathclyde is a charitable body,
registered in Scotland, number SC015263
More information about the DFTB-Plus-User