Chapter 6: Q17E (page 572)
Question: 6.17 Benchmarking is field of study that involves identifying representative workloads to run on specific computing platforms in order to be able to objectively compare performance of one system to another. In this exercise we will compare two classes of benchmarks: the Whetstone CPU benchmark and the PARSEC Benchmark suite. Select one program from PARSEC. All programs should be freely available on the Internet. Consider running multiple copies of Whetstone versus running the PARSEC Benchmark on any of systems described in Section 6.11.
6.17.1 [60] what is inherently different between these two classes of workload when run on these multi-core systems?
6.17.2 [60] In terms of the Roofline Model, how dependent will the results you obtain when running these benchmarks be on the amount of sharing and synchronization present in the workload used?
Short Answer
6.17.1
The two differences between these two suites:
Whetstone
PARSEC
6.17.2
The benchmark of the “PARSEC” should be only impacted by the synchronization and the sharing.