From cTuning.org

Navigation: cTuning.org > CTools > CBench

At the end of 2006, Grigori Fursin and Olivier Temam wanted to extend their research to enable practical iterative compilation and run-time adaptation for programs with varying behavior (FCOP2005", UNIDAPT Framework) but faced a problem of a lack of multiple datasets for available open-source benchmarks to continue experiments. Hence, they decided to create their own benchmark with multiple datasets based on the popular open-source MiBench benchmark.

In order, to simplify experimentation and enhance portability of the benchmark, Grigori decided to slightly modify the source code and directory structure of all benchmarks and added a loop wrapper around the main function to control overall execution time. This control proved convenient when executing benchmarks on servers, small embedded architectures or using architecture/system simulators.

In 2007 we decided to release the benchmark and all datasets in a hope that it would be useful to the IT R&D community. Since then the benchmark and datasets have been updated several times thanks to the feedback from the users.

In February, 2009 a considerable update has been made by Erven Rohou to make programs more C-compliant to compile them on various embedded architectures and pass them through GCC4CIL compiler. Grigori updated all the dataset scripts and provided wrappers to be used with Continuous Collective Compilation Framework. At the same time, after receiving feedback and requests about adding more programs/datasets to the benchmark, we decided to move all the developments to the new Collective Benchmark not to confuse it with the original MiBench (the first added program is bzip2 with 32 datasets) and prepared this website to collaboratively enhance and analyze it.

cBench/MiDataSets is now a community driven project and we would like to add more open-source sequential and parallel programs with multiple datasets in a hope that it will enable realistic benchmarking and research on program and architecture optimization.

Acknowledgments

Here is the list of contributors to cBench/MiDataSets:

• Erven Rohou (IRISA, France)
• Abdul Wahid Memon (UVSQ/Paris South University/INRIA, France)
• Menjato Rakoto (Paris South University and INRIA, France)
• Yang Chen (ICT, China)
• Yuanjie Huang (ICT, China)
• Chengyong Wu (ICT, China)
• Kenneth Hoste (Ghent University, Belgium)
• Veerle Desmet (Ghent University, Belgium)
• John Cavazos (University of Delaware, USA)
• Michael O'Boyle (University of Edinburgh, UK)
• Olivier Temam (INRIA Saclay, France)
• Grigori Fursin (UVSQ/INRIA, France)