Date Added: Jul 2010
The authors automatically generate an nVidia parallel CUDA graphics card kernel with identical functionality to existing highly optimised ancient sequential C code. Essentially generic GPGPU C++ code supplied by the hardware manufacturer is converted into a BNF grammar. Strongly typed genetic programming uses the BNF to generate compliable and runnable graphics card kernels, which terminate. Their fitness is given by running the population on a GPU against a test suite used to test the original sequential code.