CFD parallel processing method for thermal hydraulic simulation software of nuclear reactor

Abstract

According to the CFD parallel processing method for the thermal hydraulic simulation software of the nuclear reactor, by means of the SW many-core architecture, the CFD simulation time can be greatlyshortened, and the program running performance is improved. The method comprises the steps that the value of n2 in matrix multiplication in thermal hydraulic simulation software CFD is judged, the matrix multiplication represents that A, B and C represent double-precision floating point matrixes, n1, n2 and n3 represent the sizes of matrix dimensions, N is equal to n2, and N represents the order of a spectral unit; if 12 < = n2 < = 24, whether n1 is equal to N2 or not is judged; if n1-N2, data segmentation is conducted on the matrix A according to the number M of the slave cores, the segmenteddata are distributed to the local memories of the M slave cores, the matrix B is completely read into the local memories of the M slave cores, each slave core executes a matrix multiplication calculation task according to the data in the local memories, and after the task is completed, a calculation result is distributed back to the address of the main memory corresponding to the matrix C. The method is suitable for the field of nuclear reactor thermal hydraulic simulation.

Description

technical field [0001] The invention relates to the fields of high-performance computing and nuclear science technology and engineering, in particular to a CFD parallel processing method for nuclear reactor thermal-hydraulic simulation software. Background technique [0002] At present, a heterogeneous new processor composed of a general-purpose CPU and a dedicated acceleration core has become the development trend of supercomputers. Sunway TaihuLight is a typical supercomputer based on new heterogeneous processors. It consists of 40,960 self-developed Sunway 26010 many-core processors. Each processor contains 4 core groups (Code Group, CG). 260 heterogeneous cores. Such as figure 1 As shown, a single core group consists of a control core / master core (Manage Processing Element, MPE) and a computing core / slave core (Computing Processing Element, CPE) array (8×8 grid structure), each slave core has 16KB L1 instruction cache and 64KB local memory, MPE and CPE share 8GB main ...

AI Technical Summary

Problems solved by technology

[0008] The technical problem to be solved by the present invention is to provide a CFD parallel processing method for nuclear reactor thermal hydraulic simulation software to solve the problem of system performance degradation in the prior art

Images