Simulation cloning on GP-GPU high performance computing systems
Computer Simulations are being used to simulate real world systems in a lot of fields. They are being used so extensively to make real time decision that it is important to get the results of the simulations as fast as possible. This has led to application of Parallel and Distributed Computing to sp...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project (FYP) |
| Language: | English |
| Published: |
2012
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/48800 |
| _version_ | 1826116070329548800 |
|---|---|
| author | Akash Jana |
| author2 | School of Computer Engineering |
| author_facet | School of Computer Engineering Akash Jana |
| author_sort | Akash Jana |
| collection | NTU |
| description | Computer Simulations are being used to simulate real world systems in a lot of fields. They are being used so extensively to make real time decision that it is important to get the results of the simulations as fast as possible. This has led to application of Parallel and Distributed Computing to speed them up. In this project, a simulation cloning system on GPGPU cluster is implemented and tested.
GPUs have emerged as High Performance Computing hardware in the past few years. They offer speedup through multi-core processors and light weight threads. In this project, the simulation is implemented to run only on GPU. The speedup obtained by use of GPU has been recorded.
Simulation cloning is an interactive technique which clones the simulation at decision point to explore what difference a change in parameters would have made at that point. Simulation cloning is used to save execution time as the clone and its parent simulation have shared an execution path. It is a better solution to trivially running the simulation again with different set of parameters. Simulation cloning with Distributed Computing offers to speed up the whole execution of simulation and explore all possible execution paths. Such a mechanism has been implemented on top of GPUs in this project.
Thus, dual level of speeding up measures has been implemented in the project. The report contains the details of the implementation. The system has also been benchmarked and the results have been recorded. Although, the system offers fast runtimes, a state-space pruning heuristic for Forest fire simulation would further speedup the whole system. |
| first_indexed | 2024-10-01T04:05:24Z |
| format | Final Year Project (FYP) |
| id | ntu-10356/48800 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T04:05:24Z |
| publishDate | 2012 |
| record_format | dspace |
| spelling | ntu-10356/488002023-03-03T20:29:32Z Simulation cloning on GP-GPU high performance computing systems Akash Jana School of Computer Engineering Parallel and Distributed Computing Centre Michael Harold Lees DRNTU::Engineering::Computer science and engineering::Computer systems organization::Performance of systems Computer Simulations are being used to simulate real world systems in a lot of fields. They are being used so extensively to make real time decision that it is important to get the results of the simulations as fast as possible. This has led to application of Parallel and Distributed Computing to speed them up. In this project, a simulation cloning system on GPGPU cluster is implemented and tested. GPUs have emerged as High Performance Computing hardware in the past few years. They offer speedup through multi-core processors and light weight threads. In this project, the simulation is implemented to run only on GPU. The speedup obtained by use of GPU has been recorded. Simulation cloning is an interactive technique which clones the simulation at decision point to explore what difference a change in parameters would have made at that point. Simulation cloning is used to save execution time as the clone and its parent simulation have shared an execution path. It is a better solution to trivially running the simulation again with different set of parameters. Simulation cloning with Distributed Computing offers to speed up the whole execution of simulation and explore all possible execution paths. Such a mechanism has been implemented on top of GPUs in this project. Thus, dual level of speeding up measures has been implemented in the project. The report contains the details of the implementation. The system has also been benchmarked and the results have been recorded. Although, the system offers fast runtimes, a state-space pruning heuristic for Forest fire simulation would further speedup the whole system. Bachelor of Engineering (Computer Science) 2012-05-10T00:48:03Z 2012-05-10T00:48:03Z 2012 2012 Final Year Project (FYP) http://hdl.handle.net/10356/48800 en Nanyang Technological University 51 p. application/pdf |
| spellingShingle | DRNTU::Engineering::Computer science and engineering::Computer systems organization::Performance of systems Akash Jana Simulation cloning on GP-GPU high performance computing systems |
| title | Simulation cloning on GP-GPU high performance computing systems |
| title_full | Simulation cloning on GP-GPU high performance computing systems |
| title_fullStr | Simulation cloning on GP-GPU high performance computing systems |
| title_full_unstemmed | Simulation cloning on GP-GPU high performance computing systems |
| title_short | Simulation cloning on GP-GPU high performance computing systems |
| title_sort | simulation cloning on gp gpu high performance computing systems |
| topic | DRNTU::Engineering::Computer science and engineering::Computer systems organization::Performance of systems |
| url | http://hdl.handle.net/10356/48800 |
| work_keys_str_mv | AT akashjana simulationcloningongpgpuhighperformancecomputingsystems |