ECS Logo

ISSN: 1110-2586

Genetic Programming Scheme for Optimizing Register Allocation

Mostafa Sami
Computer Science Department, Faculty of Computers and Information, Helwan University,Cairo, Egypt
Mohamed Haggag
Computer Science Department, Faculty of Computers and Information, Helwan University,Cairo, Egypt
Heba Thabit
Computer Science Department, Faculty of Computers and Information, Helwan University,Cairo, Egypt
Abstract
Code execution is primarily dependent on the strategy used in register allocation. In this paper, a new register allocation scheme based on Genetic programming is proposed. Evaluating the performance is achieved by monitoring the code execution time and spilling cost. A seed priority function along with a list of random expressions, constructed from basic blocks’ properties, provide the initial population. Successive generations undergo some GP operators so that the fitness of individuals, representing generated priority functions, is enhanced from generation to another. Results proved that the proposed scheme optimizes the allocation process by accelerating the overall execution time and minimizing the spilling cost.
Keywords
Genetic programming, Register allocation, Spilling, Compiler optimization