Genetic Programming Applied to Compiler Heuristic Optimization

Created by W.Langdon from gp-bibliography.bib Revision:1.8051

@InProceedings{stephenson03,

• author = "Mark Stephenson and Una-May O'Reilly and Martin C. Martin and Saman Amarasinghe",
• title = "Genetic Programming Applied to Compiler Heuristic Optimization",
• booktitle = "Genetic Programming, Proceedings of EuroGP'2003",
• year = "2003",
• editor = "Conor Ryan and Terence Soule and Maarten Keijzer and Edward Tsang and Riccardo Poli and Ernesto Costa",
• volume = "2610",
• series = "LNCS",
• pages = "238--253",
• address = "Essex",
• publisher_address = "Berlin",
• month = "14-16 " # apr,
• organisation = "EvoNet",
• publisher = "Springer-Verlag",
• keywords = "genetic algorithms, genetic programming, SBSE",
• ISBN = "3-540-00971-X",
• URL = "http://www.martincmartin.com/papers/GPApppliedToCompilerHeuristicOptimizationStephenson.pdf",
• URL = "http://www.cag.lcs.mit.edu/metaopt/papers/metaopt-euro-gp.ps",
• URL = "http://citeseer.ist.psu.edu/574999.html",
• URL = "http://citeseer.ist.psu.edu/569364.html",
• DOI = "doi:10.1007/3-540-36599-0_22",
• abstract = "Genetic programming (GP) has a natural niche in the optimization of small but high payoff software heuristics. We use GP to optimize the priority functions associated with two well known compiler heuristics: predicated hyperblock formation, and register allocation. Our system achieves impressive speedups over a standard baseline for both problems. For hyperblock selection, application-specific heuristics obtain an average speedup of 23percent (up to 73percent) for the applications in our suite. By evolving the compiler's heuristic over several benchmarks, the best general-purpose heuristic our system found improves the predication algorithm by an average of 25percent on our training set, and 9percent on a completely unrelated test set. We also improve a well-studied register allocation heuristic. On average, our system obtains a 6percent speedup when it specializes the register allocation algorithm for individual applications. The general-purpose heuristic for register allocation achieves a 3percent improvement.",
• notes = "EuroGP'2003 held in conjunction with EvoWorkshops 2003",

}

Genetic Programming entries for Mark Stephenson Una-May O'Reilly Martin C Martin Saman Amarasinghe

Citations