SUBJECT: Ph.D. Proposal Presentation
BY: Kyle Remley
TIME: Monday, November 16, 2015, 1:00 p.m.
PLACE: Boggs, 3-47
TITLE: Development of Methods for High-Performance Computing Implementations of the Deterministic Stage of COMET Calculations
COMMITTEE: Dr. Farzad Rahnema, Chair (NRE)
Dr. Dingkang Zhang (NRE)
Dr. Bojan Petrovic (NRE)
Dr. Tom Morley (MATH)
Dr. Alireza Haghighat (VT)


The Coarse Mesh Radiation Transport (COMET) method is a radiation transport solver that has been used to solve whole core reactor eigenvalue and flux distribution problems. The method has been benchmarked against many different reactor types (e.g., PWR, BWR, HTGR, ABTR). A strength of the method is its formidable accuracy and computational efficiency: COMET solutions are computed to Monte Carlo accuracy in a runtime that is several orders of magnitude faster than stochastic calculations. However, with the growing ubiquity of parallel machines (e.g., gpu systems), serial implementations of COMET calculations will become less desirable, as parallel machines show more promise for maintaining and increasing computational efficiency in future applications.
It is under this motivation that methods for a parallel execution of the deterministic COMET calculations will be developed. COMET involves inner and outer iterations. The inner iterations involve sweeps where calculations for each mesh are independent during each iteration, so the algorithm is amenable to parallelization. Multiple decompositions will be explored to help achieve an optimal speedup with a parallel implementation of the COMET solution algorithm. In addition, applications of previously used acceleration methods will be investigated.