Retargeting Sequential Image-Processing Programs for Data Parallel Execution
IEEE Transactions on Software Engineering
Estimating critical region parallelism to guide platform retargeting
Proceedings of the 43rd annual Southeast regional conference - Volume 1
Scientific applications vs. SPEC-FP: a comparison of program behavior
Proceedings of the 20th annual international conference on Supercomputing
| Hi-index | 0.00 |
Scientific, symbolic, and multimedia applicationspresent diverse computing workloads with differenttypes of inherent parallelism. Tomorrow's processorswill employ varying combinations of parallel executionmechanisms to efficiently harness this parallelism. Theexplosion of consumer products that incorporate highperformance embedded computing will increase thestratification of the processor design space. However,existing code assets are limited to sequentialexpression of what should be highly parallelalgorithms. Retargeting to parallel mechanisms isdifficult, but can provide significant increases inefficiency. It is desirable to estimate potentialparallelism before undertaking the expensive processof reverse engineering and retargeting. This paperpresents a lightweight dynamic analysis technique forcharacterizing the types of parallelism that areinherent in a given program to estimate the potentialbenefit of retargeting. The technique is validated onSpec95 and MediaBench benchmarks widely used toevaluate processor performance. Results correlate wellwith previous experience in parallelizing these well-understoodapplications.