Loading...
Please wait, while we are loading the content...
Similar Documents
Parallelization of calculations using GPU in optimization approach for macromodels construction
| Content Provider | Semantic Scholar |
|---|---|
| Author | Stakhiv, Petro Strubytska, Iryna Kozak, Yuriy G. |
| Copyright Year | 2012 |
| Abstract | Construction of mathematical models for nonlinear dynamical systems using optimization requires significant computation efforts to solve the optimization task. The most CPU time is required by optimization procedure for goal function calculations, which is repeated many times for different model parameters. This allows to use processors with SIMD architecture of calculation parallelization. The effectiveness of such parallelization is the subject of investigation in this paper. Problem Statement The process of design and analysis of modern dynamical systems with large number of components of different nature requires significant computation resources. It is caused by considerable number of components of the system being designed and variety of physical phenomena to be taken into account. The usage of macromodels in such conditions allows significant reducing of required computation efforts because it makes it possible to ignore not important effects for particular analysis. Macromodels can be used to describe single components as well as subsystems of significant size including elements of different nature. Such state of the problem to be considered leads to the necessity to develop universal approaches intended for construction of complex dynamical object macromodels in the form useful for their further analysis There are many approaches and methods which might be used to construct macromodels of nonlinear dynamical systems but their usage is limited due to complexity of the problem. Therefore the task to develop an universal approach for nonlinear dynamical macromodels construction is still not solved. One promising approach, which has enough universality, is the use of optimization. But this approach has a significant disadvantage: the optimization task to be solved in this approach is very complex and often requires too big computation efforts for solution. So techniques for simplification of this optimization task are being developed (1). One more direction which allows to reduce the time needed to solve optimization task using parallelization based on processors with SIMD architecture is considered in this paper. Processors with SIMD architecture have much better ratio performance/price then regular computer processors. Authors used graphical processor NVIDIA GeForce GTS250, 1024 Mb which has SIMD architecture to verify the approach. |
| Starting Page | 7 |
| Ending Page | 9 |
| Page Count | 3 |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://pe.org.pl/articles/2012/3a/3.pdf |
| Alternate Webpage(s) | http://ena.lp.edu.ua:8080/bitstream/ntb/22485/1/27-Stakhiv-30.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
