Jump to: navigation, search
Navigation: > CTools > CCC

Automating compiler and architecture design and program optimization using iterative feedback directed compilation, statistical analysis and machine learning requires performing a very large amount of experiments. During the MHAOTEU project(1999-2002) Grigori Fursin started developing first collaborative tools to automate iterative compilation together with his source-to-source compiler to find good loop transformations and their parameters (FUR2004, EOS Framework). In 2005 this framework evolved into the FCO Framework to move away from the unstable source-to-source transformation tools and reuse production quality compilers as a research infrastructure. It was combined with the newly developed Interactive Compilation Interface for PathScale/Open64 Interactive Compilation Interface and supported different search strategies to select select fine-grain optimization and parameters such as loop tiling, unrolling, prefetching, loop distribution and fusion, array padding, vectorization, etc. This framework has been used in a number of international projects. Since 2005, based on its pros and cons, Grigori started developing a new Continuous Collective Compilation Framework within the MILEPOST project and decided to make it publicly available in 2007. This framework has been used to build a large training set (more than 5,000,000 experiments in 2 years) to create a machine learning model that can predict good program optimizations based on program static and dynamic (hardware counters) features. At the end of 2008, Grigori rewrote it completely to enable third-party plugins and common API to work with the Collective Optimization Database and provide different data analysis and optimization plugins. You are welcome to join the project and extend the CCC core framework or add new plugins.


Here is the list of CCC contributors/evaluators:

  • Yuriy Kashnikov (UVSQ, France)
  • Abdul Wahid Memon (UVSQ, France)
  • Menjato Rakoto (Paris South University and INRIA, France)
  • Yuanjie Huang (ICT, China)
  • Mircea Namolaru (IBM Haifa, Israel)
  • John Thomson (University of Edinburgh, UK)
  • Phil Barnard (ARC, UK)
  • Erven Rohou (IRISA, France)
  • Mingjie Xing (ICT, China)
  • Grigori Fursin (UVSQ/INRIA, France)
Locations of visitors to this page