organised by Prof. J. Magiera
Cracow University of Technology, Krakow, Poland
In the last few years an important tendency in the microprocessor industry has been observed, tightly related to the fact that the currently dominating silicon based semiconductor technology had almost reached its physical limits. Having no chance to pack more clock cycles per second, the manufacturers decided to implement advanced concepts that evolved from implicit type parallel architectures like e.g. pipelined/superscalar processors, multi-threading/simultaneous multithreading processors to explicit type parallel architectures called as many-core or multi-core processors. Currently, even some the netbooks are being built around such processors, viz. Intel Atom 330, a dual-core, hyperthreading enabled processer, visible in OS-es as a four way system.
For the engineering community this ubiquitous parallel architecture revolution is a true challenge as it demands a new class of software. It is no longer that only the specialty software crafted for the specialty systems has to be designed/implemented within the realms of parallel computing paradigms. And it is no longer that only those premium-price-tag systems are capable of delivering power adequate for solving real engineering problems. The wealth of professional engineering software packages offered today for the PC platform is perhaps the best proof of it. But the abundance of many-core offerings, various microarchitecture implementations, proprietary solutions, different cache sizes and memory management algorithms, various extensions, etc. do not make this move easy. This is especially true in case of sometimes extremely valuable legacy software designed in the sequential algorithms era which is still very often used in the scientific/engineering community. The fact that in many cases the programmers who are to work on porting the legacy code are not its authors also adds to complexities of this task.
The aim of this session is to provide a forum to discuss the problems of porting/refactoring the classic software to the parallel computing world and share the best strategies worked out in this important assets preserving area.
- Tools and frameworks for automatic parallelization of programs on multithreading/multi-core systems
- Strategies for manual refactoring of the legacy codes
- Libraries for parallelization of the legacy code
- Instruction level parallelism/Message passing algorithms
- Debugging, testing and profiling strategies
- Successful implementations of legacy codes
Please note that papers that are found to fall outside the scope of
this session may be considered for other sessions.
To submit abstracts for this special session, please email the organisers directly at (plmagier AT cyf-kr.edu.pl) or alternatively submit your abstract directly to the Conference Editor using the abstract submission portal. If you use the portal, please do not forget to mention that your abstract is part of special session PARENG2011-S10.