Evolving GPU Machine Code
| Autor: | Pereira Da Silva, Cleomar, Mota Dias, Douglas, Bentes, Cristiana, Cavalcanti Pacheco, Marco Aurelio, Fontoura Cupertino, Leandro |
|---|---|
| Přispěvatelé: | Pontifícia Universidade Católica do Rio de Janeiro (PUC-Rio), Instituto Federal Catarinense (IFC), Universidade do Estado do Rio de Janeiro [Rio de Janeiro] (UERJ), Système d’exploitation, systèmes répartis, de l’intergiciel à l’architecture (IRIT-SEPIA), Institut de recherche en informatique de Toulouse (IRIT), Université Toulouse 1 Capitole (UT1), Université Fédérale Toulouse Midi-Pyrénées-Université Fédérale Toulouse Midi-Pyrénées-Université Toulouse - Jean Jaurès (UT2J)-Université Toulouse III - Paul Sabatier (UT3), Université Fédérale Toulouse Midi-Pyrénées-Centre National de la Recherche Scientifique (CNRS)-Institut National Polytechnique (Toulouse) (Toulouse INP), Université Fédérale Toulouse Midi-Pyrénées-Université Toulouse 1 Capitole (UT1), Université Fédérale Toulouse Midi-Pyrénées, National Counsel of Technological and Scientific Development (CNPq), Carlos Chagas Filho Research SupportFoundation (FAPERJ), Grélaud, Françoise |
| Jazyk: | angličtina |
| Rok vydání: | 2015 |
| Předmět: | |
| Zdroj: | Journal of Machine Learning Research Journal of Machine Learning Research, Microtome Publishing, 2015, 16, pp.673--712 |
| ISSN: | 1532-4435 1533-7928 |
| Popis: | International audience; Parallel Graphics Processing Unit (GPU) implementations of GP have appeared in the literature using three main methodologies: (i) compilation, which generates the individuals in GPU code and requires compilation; (ii) pseudo-assembly, which generates the individuals in an intermediary assembly code and also requires compilation; and (iii) interpretation, which interprets the codes. This paper proposes a new methodology that uses the concepts of quantum computing and directly handles the GPU machine code instructions. Our methodology utilizes a probabilistic representation of an individual to improve the global search capability. In addition, the evolution in machine code eliminates both the overhead of compiling the code and the cost of parsing the program during evaluation. We obtained up to 2.74 trillion GP operations per second for the 20-bit Boolean Multiplexer benchmark. We also compared our approach with the other three GPU-based acceleration methodologies implemented for quantum-inspired linear GP. Significant gains in performance were obtained. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|