A new prototype (or mini-app) PATMOS for Monte Carlo neutron transport is under development at CEA [1]. PATMOS is complex enough to be representative of a real simulation code and at the same time conceived to be easy to change and adapt. This prototype is devoted to the testing of algorithms for high performance computations on different modern architectures. The final goal is to perform pin-by-pin full core depletion calculations for large nuclear power reactors with realistic temperature fields. 

​PATMOS is entirely written in C++11 but meant to be used in Python, via a SWIG-generated interface, for all user accessible objects. It is object oriented and makes heavy use of polymorphism in order to always allow the choice between competing algorithms: as an example, in PATMOS one can mix nuclides with pre-computed Doppler-broadened cross sections and nuclides with on-the-fly Doppler broadening.

Prototyping of PATMOS is first performed in Python for agile programming. So there actually are two versions of PATMOS, one in Python and one in C++. Verification tests assure that the two versions give the same results.

Testing PATMOS performances for the Hoogenboom-Martin benchmark at the TGCC supercomputer.

References

[1] E. Brun, S. Chauveau, F. Malvagi, PATMOS: A prototype Monte Carlo transport code to test high performance architectures, in Proc. M&C 2017, Jeju, Korea, April 16-20, 2017.