研究目的
To develop parallel domain decomposition methods for a quantum-corrected drift–diffusion (QCDD) model for MOSFET devices, focusing on hybrid MPI/OpenMP parallelization algorithms suitable for many-core architectures.
研究成果
The developed hybrid MPI/OpenMP parallelization methods for the QCDD model show significant improvements in parallel performance, especially when combined with the splitting-up operator method for intranode parallelization. The RAS method extensions for internode parallelization demonstrate good scalability, particularly when subdomains are homogeneous. The proposed methods are suitable for parallel computers with multi- or many-core architectures.
研究不足
The study is limited to specific architectures (NEC SX-ACE and Cray XC40) and a particular model (three-dimensional Si bulk n-MOSFET). The parallel efficiency is affected by the number of domain decompositions and the boundary conditions.
