
RPPM : Rapid Performance Prediction of Multithreaded workloads on multicore processors
- Author
- Sander De Pestel, Sam Van den Steen, Shoaib Akram and Lieven Eeckhout (UGent)
- Organization
- Abstract
- Analytical performance modeling is a useful complement to detailed cycle-level simulation to quickly explore the design space in an early design stage. Mechanistic analytical modeling is particularly interesting as it provides deep insight and does not require expensive offline profiling as empirical modeling. Previous work in mechanistic analytical modeling, unfortunately, is limited to single-threaded applications running on single-core processors. This work proposes RPPM, a mechanistic analytical performance model for multi-threaded applications on multicore hardware. RPPM collects microarchitecture-independent characteristics of a multi-threaded workload to predict performance on a previously unseen multicore architecture. The profile needs to be collected only once to predict a range of processor architectures. We evaluate RPPM's accuracy against simulation and report a performance prediction error of 11.2% on average (23% max). We demonstrate RPPM's usefulness for conducting design space exploration experiments as well as for analyzing parallel application performance.
- Keywords
- DESIGN-SPACE, REGRESSION, FRAMEWORK, POWER, MODEL
Downloads
-
ispass2019-sander.pdf
- full text
- |
- open access
- |
- |
- 1.18 MB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-8625897
- MLA
- De Pestel, Sander, et al. “RPPM : Rapid Performance Prediction of Multithreaded Workloads on Multicore Processors.” 2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), 2019, pp. 257–67, doi:10.1109/ispass.2019.00038.
- APA
- De Pestel, S., Van den Steen, S., Akram, S., & Eeckhout, L. (2019). RPPM : Rapid Performance Prediction of Multithreaded workloads on multicore processors. 2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), 257–267. https://doi.org/10.1109/ispass.2019.00038
- Chicago author-date
- De Pestel, Sander, Sam Van den Steen, Shoaib Akram, and Lieven Eeckhout. 2019. “RPPM : Rapid Performance Prediction of Multithreaded Workloads on Multicore Processors.” In 2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), 257–67. https://doi.org/10.1109/ispass.2019.00038.
- Chicago author-date (all authors)
- De Pestel, Sander, Sam Van den Steen, Shoaib Akram, and Lieven Eeckhout. 2019. “RPPM : Rapid Performance Prediction of Multithreaded Workloads on Multicore Processors.” In 2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), 257–267. doi:10.1109/ispass.2019.00038.
- Vancouver
- 1.De Pestel S, Van den Steen S, Akram S, Eeckhout L. RPPM : Rapid Performance Prediction of Multithreaded workloads on multicore processors. In: 2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS). 2019. p. 257–67.
- IEEE
- [1]S. De Pestel, S. Van den Steen, S. Akram, and L. Eeckhout, “RPPM : Rapid Performance Prediction of Multithreaded workloads on multicore processors,” in 2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), Madison, WI, 2019, pp. 257–267.
@inproceedings{8625897, abstract = {{Analytical performance modeling is a useful complement to detailed cycle-level simulation to quickly explore the design space in an early design stage. Mechanistic analytical modeling is particularly interesting as it provides deep insight and does not require expensive offline profiling as empirical modeling. Previous work in mechanistic analytical modeling, unfortunately, is limited to single-threaded applications running on single-core processors. This work proposes RPPM, a mechanistic analytical performance model for multi-threaded applications on multicore hardware. RPPM collects microarchitecture-independent characteristics of a multi-threaded workload to predict performance on a previously unseen multicore architecture. The profile needs to be collected only once to predict a range of processor architectures. We evaluate RPPM's accuracy against simulation and report a performance prediction error of 11.2% on average (23% max). We demonstrate RPPM's usefulness for conducting design space exploration experiments as well as for analyzing parallel application performance.}}, author = {{De Pestel, Sander and Van den Steen, Sam and Akram, Shoaib and Eeckhout, Lieven}}, booktitle = {{2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS)}}, isbn = {{9781728107462}}, keywords = {{DESIGN-SPACE,REGRESSION,FRAMEWORK,POWER,MODEL}}, language = {{eng}}, location = {{Madison, WI}}, pages = {{257--267}}, title = {{RPPM : Rapid Performance Prediction of Multithreaded workloads on multicore processors}}, url = {{http://dx.doi.org/10.1109/ispass.2019.00038}}, year = {{2019}}, }
- Altmetric
- View in Altmetric
- Web of Science
- Times cited: