Predicting Inter-Thread Cache Contention on a Chip Multi-Processor Architecture
HPCA '05 Proceedings of the 11th International Symposium on High-Performance Computer Architecture
Pin: building customized program analysis tools with dynamic instrumentation
Proceedings of the 2005 ACM SIGPLAN conference on Programming language design and implementation
Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)
Discovering and Exploiting Program Phases
IEEE Micro
Addressing shared resource contention in multicore processors via scheduling
Proceedings of the fifteenth edition of ASPLOS on Architectural support for programming languages and operating systems
AKULA: a toolset for experimenting and developing thread placement algorithms on multicore systems
Proceedings of the 19th international conference on Parallel architectures and compilation techniques
Performance Prediction on Multi-core Processors
CICN '10 Proceedings of the 2010 International Conference on Computational Intelligence and Communication Networks
A Machine Learning Based Meta-Scheduler for Multi-Core Processors
International Journal of Adaptive, Resilient and Autonomic Systems
| Hi-index | 0.00 |
Programs co-running on cores share resources on multi-core processor systems. It is now well known that interference between the programs arising from the sharing may result in severe performance degradations. It is the objective of recent research in system scheduling to be aware of shared resource requirements of the running programs (threads). To this end AKULA is a toolset recently developed that provides a platform for experiments and developing thread scheduling algorithms on multi-core processors. In AKULA a bootstrapping module works on the basis of previously collected performance data of programs to simulate program execution on multi-cores. In this paper we describe a different approach where that augments such a bootstrapping module with a model built using machine learning techniques. The proposed model will extend the bootstrapping module's ability to predict degradation in performance due to sharing where previous performance data is not available for pairing /co-scheduling of applications. Also the proposed approach allows greater scalability for variable number of processor cores sharing the resources.