Synchronization migration for performance enhancement in a DOACROSS loop
An efficient technique for migrating the synchronization operations is proposed. This technique rewrites the original statements, Send_Signal(S) to be moved up and Wait_Signal(S, i−d) to be moved down, or rearranges the sequence of statements or the position of array element in a synchronization region by data dependence analysis to migrate the synchronization operations. Theorems show that the migration of synchronization operation can reduce the parallel execution time of loop significantly. Perfect benchmarks are employed to measure the system performance after migration. Experimental result shows that the enhancement is very significant.
KeywordsCode migration Data dependence LBD LFD Synchronization Synchronization region
Unable to display preview. Download preview PDF.
- 1.C. Q. Zhu and P. C. Yew, ‘A scheme to enforce data dependence on large multiprocessor systems”, IEEE trans. Software Eng. June 1987, 726–739.Google Scholar
- 2.D. K. Chen and P. C. Yew, ‘An empirical study on DOACROSS loops', Proc. of supercomputing. November 1991, 18–22.Google Scholar
- 3.H. M. Su, and P. C Yew, ‘On data synchronization for multiprocessors', Proceeding l 6th Annual International Symposium on Computer Architecture, May 1989, 416–423.Google Scholar
- 4.J. R. Allen and K. Kennedy, ‘Automatic translation of FORTRAN programs to vector form', ACM trans. Programming Language and Syst, October 1987, 491–542.Google Scholar
- 5.M. Wolfe, ‘Multiprocessor synchronization for concurrent loops’, IEEE Software, January 1988, 34–42.Google Scholar
- 6.P. Tang, P. C. Yew and C. Q. Zhu, ‘Compiler techniques for data synchronization in nested parallel loops', Proc. 1990 International Conference on Supercomputing, Amsterdam, Holland, May 1990, 177–186Google Scholar
- 7.R. Eigenmann, J. Hoeflinger, Z. Li, and D. Padua, ‘Experiences in the automatic parallelization of four perfect-benchmark programs', Proceeding of the Fourth Workshop on Languages and Compilers for Parallel Computing. August 1991, pp. 87–95.Google Scholar
- 8.Rong-Yuh Hwang and Feipei Lai, ‘An Intelligent Code Migration Technique for Synchronization Operations on a Multiprocessor’, IEE Proceedings-E Computers and Digital Techniques, March 1995, pp107–116.Google Scholar
- 9.Rong-Yuh Hwang and Feipei Lai, ‘Intelligent Code Migration Technique for Synchronization Operations on a Multiprocessor', International Symposium on a Parallel Archilectures, Algorithms, and Networks, December 1994, pp121–127.Google Scholar
- 10.S. P. Midkiff and D. A. Padua, ‘Compiler algorithm for synchronization', IEEE trans. Comput, C-36, 12, December 1987, pp1485–1495.Google Scholar
- 11.Z. Li, ‘Compiler algorithms for event variable synchronization', Proc. 1991 ACM International Conference on Supercomputing, July 1991, pp85–95.Google Scholar