Portable and Predictable Performance on Het Embedded Systems
| Reference number | |
| Coordinator | SICS SWEDISH ICT AB |
| Funding from Vinnova | SEK 4 852 482 |
| Project duration | September 2012 - October 2015 |
| Status | Completed |
Purpose and goal
In order to remain competitive in the fast changing global market, embedded system manufacturers need in particular to decrease hardware and energy costs of their offerings. The current industrial practice for determining system configurations involves manual performance estimations, which becomes ever harder with new complex hardware designs. The PaPP project aimed at enabling software with predictable performance and reduced resource usage for current and future parallel platforms.
Expected results and effects
SICS developed a practical performance prediction technique and a prototype tool for OpenMP task-based applications. The tool allows to both exploit the hardware design space, and also guide the software optimization process. The tool has been evaluated on industrial applications from three problem domains: telecommunications, multimedia, and space, with the mean prediction accuracy of 18%. The gained expertise will be exploited in R&D projects with the Swedish industry.
Planned approach and implementation
The PaPP performance prediction technique is enabled by the task-based programming model, and the implementation is based on the public-domain OpenMP 3.0 infrastructure. Performance models of target platforms are automatically derived using special ´characterization programs´, and performance models of software to be analyzed capture task performance properties and task inter-dependencies. The performance prediction tool simulates task execution on a multicore chip using the aforementioned models.