PARSECSs

PARSECSs is a suite of benchmark applications for parallel architectures.

Institution:

Research Group:

BSC Group: Computer Sciences

Researcher/s:

Dimitrios Chasapis

Description:

PARSECSs extends the PARSEC benchmark suite with task-based versions that use advanced parallelism and data-flow concepts. The project re-engineered PARSEC applications using task-based programming models and dataflow dependencies, primarily through the OmpSs/OpenMP tasking model. The work was designed as a research benchmark suite for studying programmability, scalability, runtime overheads, and performance portability of task-based execution models compared with conventional thread-based implementations. The project demonstrated that task-based programming can significantly simplify parallel application development while improving scalability in applications with complex synchronization and pipeline behavior.

Value Proposition:

Portable task-based implementation of the PARSEC benchmarks for evaluation programmability, scalability and runtime performance on multicore systems

Aplication areas:

Research/Evaluation of task-based models and runtimes, multicore architecture research, university teaching of task parallelism

Target market:

Runtime System and Compiler developers; Computer Architects; Super-computing centers and academic laboratories

Protection:

BSD License

More information

if you want to know more about this project do not hesitate to contact us