Title :
Elastic-plastic flow simulation using the Supercomputer Toolkit
Author :
Goikhman, Alexander ; Katzenelson, Jacob
Author_Institution :
Dept. of Electr. Eng., Technion-Israel Inst. of Technol., Haifa, Israel
Abstract :
This article describes the first application program run on Technion´s Supercomputer Toolkit. The program simulates the behavior of a high speed metal projectile colliding with a rigid wall. The velocity of the projectile is high enough to cause plastic deformations. After briefly describing the Toolkit and the elastic-plastic flow problem this article is concerned with the relation between the Toolkit architecture and the properties of the elastic-plastic simulation. The main conclusions are that the parallelization process is straightforward, and the obtained speedup, for a small number of processors, is about equal to the number of processors. The first item is a result of the elastic-plastic simulation computational structure, which can be viewed as computations done within cells that communicate mainly with their nearest neighbors. This structure is parallelized easily and maps simply and readily to a Toolkit parallel computing network. The computation structure leads to a fair amount of local communication among Toolkit processors, but to only a modest amount of global communication. The Toolkit´s local communication is exceedingly fast and the time for local communication is independent of the network size, which explains the second conclusion and fits the requirement of large elastic-plastic problems
Keywords :
digital simulation; flow simulation; parallel machines; parallel programming; physics computing; plastic deformation; software performance evaluation; special purpose computers; Supercomputer Toolkit; Technion; communication time; computation structure; elastic-plastic flow simulation; global communication; high speed metal projectile simulation; local communication; nearest neighbor; parallel computing network; plastic deformation; projectile collision; projectile velocity; rigid wall; special purpose computers; speedup; Computational modeling; Computer architecture; Deformable models; Jacobian matrices; Nearest neighbor searches; Parallel processing; Partial differential equations; Plastics; Projectiles; Supercomputers;
Conference_Titel :
Parallel Architectures and Compilation Techniques, 1996., Proceedings of the 1996 Conference on
Conference_Location :
Boston, MA
Print_ISBN :
0-8186-7633-7
DOI :
10.1109/PACT.1996.552659