dc.contributor.authorChandrasehar Rajaseharan.
dc.description.abstractThe proposal of refactoring applications mainly came about due to sequential algorithms that have room for improvement in terms of performance. Most sequential algorithms like matrix multiplication where the complexity is O(n3) can affect the overall execution time of a program when their size increases. This can be improved if algorithms are transformed in such a way where the operations involving the array(s) in loops can be parallelized. In the context of the project, this parallelism is mainly achieved by two techniques, Array slicing and Fortran’s intrinsic functions in Fortran 90 where the former uses the technique of vectorization while the latter uses compiler-defined algorithms to parallelize array operations. A plugin developed using Eclipse and the Photran framework implements this process by identifying loops within programs where their array operations are refactored to either array sliced forms or Fortran’s intrinsic functions. This identification process includes data dependence analysis on the loops. Conclusively, profiling tests conducted on the implementation of these two techniques in two Fortran applications computing mathematical functions such as derivatives of a function and plotting a linear curve for Millikan experimental data highlight a maximum speedup of 2.99 as the size of the vectors involved in the application increase from 100 to 100000 for the first application and a maximum speedup of 1.73 as the size of the vectors increase from 20000 to 100000 for the second application.en_US
dc.format.extent51 p.en_US
dc.rightsNanyang Technological University
dc.subjectDRNTU::Engineering::Computer science and engineering::Computer systems organization::Performance of systemsen_US
dc.titleAutomatic refactoring of Fortran applicationsen_US
dc.typeFinal Year Project (FYP)en_US
dc.contributor.supervisorStephen John Turneren_US
dc.contributor.schoolSchool of Computer Engineeringen_US
dc.description.degreeBachelor of Engineering (Computer Science)en_US
dc.contributor.organizationA*STAR Institute of High Performance Computing (IHPC)en_US
dc.contributor.researchParallel and Distributed Computing Centreen_US

Files in this item


This item appears in the following Collection(s)

Show simple item record