Session on:

Within the 2010 international Conference on
Parallel and Distributed Processing Techniques and Applications
(PDPTA'10)
Las Vegas, Nevada, USA. 12-15 July 2010

In a few years multi- and many-core processors will not just have 8, 12 or 16 cores, but many more. At the same time will the ecosystem of applications making advanced use of those architectures keep inflating at a rapid pace on supercomputers, game consoles, clusters, servers, workstations, and laptops. In this emerging combinatorial universe of architectures and applications running to a large extent in commercial environments with mixes of closed and open source systems and applications, there is a need for alternative evaluation methods to compliment e.g. code analysis and operating systems modifications.

Towards this end, this session is focused on experimentally related techniques for evaluating, comparing and determining the performance impact of running applications on multi-core computers of any kind. This includes, but is not limited to the topics listed in the Call for Papers .

---

Submission Deadline March 5 (see Important Dates)

---

 
Topics include but are not limited too:

• Benchmarking on multi-core architectures
• Bottleneck detection techniques in e.g. industrial HPC-environments/applications
• Performance evaluation with limited code and/or system access
• Experiments, evaluations or modeling of application behavior
• Experimental predictions of application behavior and estimates of approximation errors.
• Evaluations or modeling of application behavior in competitive conditions for resources (e.g. two or more applications sharing cores, memory, network, etc.)
• Tools, techniques and methods to perform experiments and analyze results/optimal configurations.
• Usage of experimental data in real world HPC environments, e.g. use in scheduling etc.