Issue No. 07 - July (1993 vol. 4)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/71.238302
<p>A new technique for estimating and understanding the speed improvement that can resultfrom executing a program on a parallel computer is described. The technique requires noadditional programming and minimal effort by a program's author. The analysis begins by tracing a sequential program. A parallelism analyzer uses information from the trace to simulate parallel execution of the program. In addition to predicting parallel performance, the parallelism analyzer measures many aspects of a program's dynamic behavior. Measurements of six substantial programs are presented. These results indicate that the three symbolic programs differ substantially from the numeric programs and, as aconsequence, cannot be automatically parallelized with the same compilation techniques.</p>
Index Termsnumeric programs; loop level parallelism; symbolic programs; speed improvement; parallel computer; sequential program; parallelism analyzer; parallel execution; parallelperformance; dynamic behavior; parallel programming; performance evaluation; programcompilers
J. Larus, "Loop-Level Parallelism in Numeric and Symbolic Programs," in IEEE Transactions on Parallel & Distributed Systems, vol. 4, no. , pp. 812-826, 1993.