Offloading Region Matching of Data Distribution Management with CUDA

Data distribution management (DDM) aims to reduce the transmission of irrelevant data between High Level Architecture (HLA) compliant simulators by taking their interesting regions into account (i.e. region matching). In a large-scale simulation, computation intensive region matching would have a direct impact on the simulation performance. To deal with the high computation cost of region matching, the whole process of region matching is offloaded to graphical processing units (GPUs) based on Computer Unified Device Architecture (CUDA). Two approaches are proposed to perform region matching in parallel. Several metrics, including different numbers of regions, different sizes of regions and different distributions of regions, are used in the experimental tests. The experimental results indicate that the performance of region matching on a GPU can be improved more than one or two orders of magnitude in comparison with that on a CPU.

[1]  Rassul Ayani,et al.  A hybrid approach to data distribution management , 2000, Proceedings Fourth IEEE International Workshop on Distributed Simulation and Real-Time Applications (DS-RT 2000).

[2]  Rassul Ayani,et al.  Optimizing cell-size in grid-based DDM , 2000, Proceedings Fourteenth Workshop on Parallel and Distributed Simulation.

[3]  IEEE Standard for Modeling and Simulation (M&S) High Level Architecture (HLA) — Framework and Rules , 2001 .

[4]  Douglas D. Wood.  Implementation of DDM in the MAK High Performance RTI , 2002 .

[5]  Rassul Ayani,et al.  Adaptive Data Distribution Management for HLA RTI , 2002 .

[6]  Bora I. Kumova.  Dynamically adaptive partition-based data distribution management , 2005, Workshop on Principles of Advanced and Distributed Simulation (PADS'05).

[7]  Jun Yu,et al.  A sort-based DDM matching algorithm for HLA , 2005, TOMC.

[8]  Azzedine Boukerche,et al.  Grid-filtered region-based data distribution management in large-scale distributed simulation systems , 2005, 38th Annual Simulation Symposium.

[9]  S.A. Manavski,et al.  CUDA Compatible GPU as an Efficient Hardware Accelerator for AES Cryptography , 2007, 2007 IEEE International Conference on Signal Processing and Communications.

[10]  Stephen John Turner,et al.  An Efficient Sort-Based DDM Matching Algorithm for HLA Applications with a Large Spatial Environment , 2007, 21st International Workshop on Principles of Advanced and Distributed Simulation (PADS'07).

[11]  Pankaj Gupta,et al.  A Comparative Study of Data Distribution Management Algorithms , 2007 .

[12]  Fatih Erdogan Sevilgen,et al.  Quadtree-based approach to data distribution management for distributed simulations , 2008, SpringSim '08.

[13]  Richard M. Fujimoto,et al.  Offloading Data Distribution Management to Network Processors in HLA-Based Distributed Simulations , 2008, IEEE Transactions on Parallel and Distributed Systems.

[14]  Yeh-Ching Chung,et al.  MGRID: a modifiable-grid region matching approach for DDM in the HLA RTI , 2009, SpringSim '09.