REPARA was presented during the recent PEGPUM 2015 workshop (LPGPU Workshop on Power Efficiente GPU and Many-core Computing) held in Amsterdam in conjunction with the HIPEAC 2015 Conference.
During this workshop, Prof. J. Daniel García presented “REPARA: Reengineering for Heterogeneous Parallelism for Performance and Energy in C++”. During this talk several results of the project were highlighted:
- The REPARA approach based on refactoring for enforcing specific device rules, apply transofrmation to specific programming models and generate FPGA code.
- To application partitioning strategy combining hardware descriptions, kernel measurements and code.
- The extensible transformation infrastructure based on a single front-end and multiple back-ends.
- The C++ attributes language developed in the context of REPARA for defining code annotations to express parallelism patterns.
The slides are available at: http://lpgpu.org/wp/wp-content/uploads/2014/09/repara.pdf.