The concrete objectives of the REPARA project are:

  • O1: Create a new abstract representation of source code powerful enough to allow code to be automatically transformed, parallelised and executed on heterogeneous computing systems as well as the associated analysis tools.
  • O2: Provide automated partitioning of sequential applications so that they can be run across parallel heterogeneous computing systems to improve performance and/or energy efficiency.
  • O3: Transform legacy or new software components to the specific programming model associated with each target processor or accelerator.
  • O4: Compile legacy or new software components for execution on reconfigurable hardware contained in a larger heterogeneous system.
  • O5: Create new predictive performance and energy efficiency models to guide source code and reconfigurable hardware transformations as well as application partitioning.
  • O6: Create a parallel heterogeneous run-time system to support the automated parallelisation and software partitioning methods used by the transformation tools to improve performance and/or energy efficiency.
  • O7: Validate the REPARA framework with a set of representative applications from the embedded and high performance computing markets.