A statistically rigorous approach for improving simulation methodology

Yi, Joshua J.; Lilja, David J.; Hawkins, Douglas M.

doi:10.1109/hpca.2003.1183546

Cited by 106 publications

(98 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…cycles per instruction, cache miss rate, branch prediction accuracy etc., on various microarchitecture configurations that offer a different mixture of bottlenecks [12][15] [16] [17]. The variation in these metrics is then used to infer the generic program behavior.…”

Section: Related Workmentioning

confidence: 99%

Measuring Program Similarity: Experiments with SPEC CPU Benchmark Suites

Phansalkar

Joshi

Eeckhout

et al. 2005

IEEE International Symposium on Performance Analysis of Systems and Software, 2005. ISPASS 2005.

125

View full text Add to dashboard Cite

show abstract

Section: Related Workmentioning

confidence: 99%

Measuring Program Similarity: Experiments with SPEC CPU Benchmark Suites

Phansalkar

Joshi

Eeckhout

et al. 2005

IEEE International Symposium on Performance Analysis of Systems and Software, 2005. ISPASS 2005.

125

View full text Add to dashboard Cite

show abstract

“…Given the importance of the problem, a fair amount of research has been done on design space exploration. Yi et al [17] use a PlackettBurman design of experiment to identify the important axes in the design space in order to drive the design process. Eyerman et al [4] leverage machine learning techniques to more quickly steer the search process to the optimum design point.…”

Section: B Design Space Explorationmentioning

confidence: 99%

“…We therefore employ a Plackett and Burman design of experiment [17] in which we systematically vary the processor configuration along different dimensions in the design space. The Plackett and Burman design of experiment involves a small number of simulations -substantially fewer simulations than simulating all possible combinations of microarchitecture parameters -while still capturing the effects of each parameter and selected interactions.…”

Section: A Model Accuracymentioning

confidence: 99%

How sensitive is processor customization to the workload's input datasets?

Breughe

Chen

et al. 2011

2011 IEEE 9th Symposium on Application Specific Processors (SASP)

View full text Add to dashboard Cite

Abstract-Hardware customization is an effective approach for meeting application performance requirements while achieving high levels of energy efficiency. Application-specific processors achieve high performance at low energy by tailoring their designs towards a specific workload, i.e., an application or application domain of interest. A fundamental question that has remained unanswered so far though is to what extent processor customization is sensitive to the training workload's input datasets. Current practice is to consider a single or only a few input datasets per workload during the processor design cycle -the reason being that simulation is prohibitively time-consuming which excludes considering a large number of datasets. This paper addresses this fundamental question, for the first time. In order to perform the large number of runs required to address this question in a reasonable amount of time, we first propose a mechanistic analytical model, built from first principles, that is accurate within 3.6% on average across a broad design space. The analytical model is at least 4 orders of magnitude faster than detailed cycle-accurate simulation for design space exploration. Using the model, we are able to study the sensitivity of a workload's input dataset on the optimum customized processor architecture. Considering MiBench benchmarks and 1000 datasets per benchmark, we conclude that processor customization is largely dataset-insensitive. This has an important implication in practice: a single or only a few datasets are sufficient for determining the optimum processor architecture when designing application-specific processors.

show abstract

“…In [24] foldover Plackett-Burman experimental designs were used to obtain a significance ordering of microarchitectural parameters. Li et al [25] presented a simple linear model for power consumption by operating system services.…”

Section: Related Workmentioning

confidence: 99%