Development of methods for high performance computing applications of the deterministic stage of comet calculations

Coarse mesh transport;Parallel computing;Reactor physics

Remley, Kyle E. ; Rahnema, Farzad Mechanical Engineering Petrovic, Bojan Zhang, Dingkang Morley, Tom Haghighat, Alireza ; Rahnema, Farzad

University:Georgia Institute of Technology

Department:Mechanical Engineering

关键词: Coarse mesh transport;    Parallel computing;    Reactor physics;   

Others  :  https://smartech.gatech.edu/bitstream/1853/58610/1/REMLEY-DISSERTATION-2016.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

The Coarse Mesh Radiation Transport (COMET) method is a reactor physics method and code that has been used to solve whole core reactor eigenvalue and flux distribution problems. A strength of the method is its formidable accuracy and computational efficiency. COMET solutions are computed to Monte Carlo accuracy on a single processor in a runtime that is several orders of magnitude faster than stochastic calculations. However, with the growing ubiquity of both shared and distributed memory parallel machines and the desire to extend the method to allow for coupling to multiphysics and on-the-fly response generation, serial implementations of COMET calculations will become less desirable. It is under this motivation that an implementation for a parallel execution of deterministic COMET calculations has been developed. COMET involves inner and outer iterations; inner iterations involve local calculations that can be carried out independently, making the algorithm amenable to parallelization. However, considerations for decomposing a problem and the distribution of data must be made. To allow for efficient parallel implementation of a distributed algorithm, changes to response data access and sweep order are made, along with considerations for communications between processors. The parallel code is implemented on several variants of the C5G7 benchmark problem to assess the scalability of the algorithm, and it is found that problems with larger numbers of coarse meshes increase the scalability of the code, which is an encouraging result. The code is further tested for full core reactor problems, where extremely efficient wall clock times (on the order of minutes) for solutions are achieved. Finally, application of the parallel code to novel implementations of COMET (e.g., problems with high flux expansions) is discussed.