Loop Fusion Methods: A Critical Review

Abstract: Contemporary applications are much more complex than before and they need more time and resources while being executed. Investigators provide various strategies to improve the process of compiling, trying to improve execution speed. Hence, distinct transformations are proposed that are frequently used in modern compilers. One of them is loop fusion in which multiple loops merge together and form an extended one. Many studies have been done on this transformation and its impact on systems performance, reflecting the importance of this issue, each of them having its own perspective in enhancing the efficiency of loop fusion. This paper demonstrates a review of prominent researches in loop fusion and assesses the advantages and disadvantages of each.

Keywords: Optimizing Compilers; Loop transformations; Loop Fusion; Data Dependence.