Lewis, T.E. and Magoulas, George D. (2011) TMBL kernels for CUDA GPUs compile faster using PTX. In: UNSPECIFIED (ed.) Proceedings of the 13th annual conference companion on Genetic and evolutionary computation - GECCO '11. New York, USA: Association for Computing Machinery, pp. 455-462. ISBN 9781450306904.
Abstract
Many of the most effective attempts to harness the power of the Graphics Processing Unit (GPU) to accelerate Genetic Programming (GP) have dynamically compiled code for individuals as they are to be evaluated. This approach executes very quickly on the GPU but is slow to compile, hence only vast data-sets fully reap its rewards. To reduce compilation time, we generate and compile code in the lower-level language PTX. We investigate this in the context of implementing Tweaking Mutation Behaviour Learning (TMBL) on the GPU. We find that for programs of 300 instructions, using PTX reduces the compile time 5.861 times and even increases the evaluation speed by 23.029%.
Metadata
| Item Type: | Book Section |
|---|---|
| School: | Birkbeck Faculties and Schools > Faculty of Science > School of Computing and Mathematical Sciences |
| Research Centres and Institutes: | Birkbeck Knowledge Lab |
| Depositing User: | Sarah Hall |
| Date Deposited: | 18 Jul 2013 12:45 |
| Last Modified: | 09 Aug 2023 12:33 |
| URI: | https://eprints.bbk.ac.uk/id/eprint/7720 |
Statistics
Additional statistics are available via IRStats2.
Archive Staff Only (login required)
