Evolving a CUDA kernel from an nVidia template

Author

Langdon, W.B. ; Harman, M.

Author_Institution

Dept. of Comput. Sci., King´´s Coll., London, UK

fYear

2010

fDate

18-23 July 2010

Firstpage

1

Lastpage

8

Abstract

Rather than attempting to evolve a complete program from scratch we demonstrate genetic interface programming (GIP) by automatically generating a parallel CUDA kernel with identical functionality to existing highly optimised ancient sequential C code (gzip). Generic GPGPU nVidia kernel C++ code is converted into a BNF grammar. Strongly typed genetic programming uses the BNF to generate compilable and executable graphics card kernels. Their fitness is given by running the population on a GPU with randomised subsets of training data itself derived from gzip´s SIR test suite. Back-to-back validation uses the original code as a test oracle.

Keywords

C language; computer graphic equipment; genetic algorithms; operating system kernels; parallel processing; software engineering; user interfaces; BNF grammar; CUDA kernel; Generic GPGPU nVidia kernel C++ code; compute unified device architecture; executable graphics card kernels; genetic interface programming; gzip; highly optimised ancient sequential C code; nVidia template; parallel CUDA kernel; Grammar; Graphics processing unit; Humans; Kernel; Testing; Training; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Evolutionary Computation (CEC), 2010 IEEE Congress on

Conference_Location

Barcelona

Print_ISBN

978-1-4244-6909-3

Type

conf

DOI

10.1109/CEC.2010.5585922

Filename

5585922