Iteration Space Slicing

 

The TRACO compiler is an implementation of loop parallelization algorithms developed by Prof. Wlodzimierz Bielecki team in the West Pomeranian University of Technology. The code of the iteration space slicing framework (ISSF) is mostly created by Marek Palkowski. The engine of transitive closure is implemented by Tomasz Klimek. The tool realizes iteration space slicing for program loops. TRACO is based on the Omega Calculator library. Loop dependence analysis is calculated by means of the Petit tool. Code generation can be realized by Barvinok/CLOOG or Omega. Details of algorithms can be found in the article below:

download

 

The software can be found here traco-0.2.tar.gz (3.9 MB) or can be downloaded with the svn repository (recommended). TRACO is currently under development and testing.

SVN is available at the sourceforge.net website
svn checkout svn://svn.code.sf.net/p/issf/code-0/trunk

http://sourceforge.net/projects/issf/

Free scheduling

 

The another way of extracting parallelism in loops. The statements can be executed parallel in a part of time when all their operands are available. The technique can extract parallel statements also for one slice in iteration space. Details can be found in the article below:

source-to-source

 

TRACO supports source-to-source transformation of C code. The output is a parallel code with OpenMP pragmas. Implementation was developed in Python.

http://sourceforge.net/projects/issf/

Example


articles

 
Show/Hide All


Citations

@article{BieleckiPK12,
  author    = {Wlodzimierz Bielecki and
               Marek Palkowski and
               Tomasz Klimek},
  title     = {Free scheduling for statement instances of parameterized
               arbitrarily nested affine loops},
  journal   = {Parallel Computing},
  volume    = {38},
  number    = {9},
  year      = {2012},
  pages     = {518-532},
  ee        = {http://dx.doi.org/10.1016/j.parco.2012.06.001},
}

@article{BeletskaBCPS11,
  author    = {Anna Beletska and
               Wlodzimierz Bielecki and
               Albert Cohen and
               Marek Palkowski and
               Krzysztof Siedlecki},
  title     = {Coarse-grained loop parallelization: Iteration Space Slicing
               vs affine transformations},
  journal   = {Parallel Computing},
  volume    = {37},
  number    = {8},
  year      = {2011},
  pages     = {479-497},
  ee        = {http://dx.doi.org/10.1016/j.parco.2010.12.005},
}

Best Paper Award

 

The paper 'Impact of Variable Privatization on Extracting Synchronization-Free Slices' has got the Best Paper Award (Theory) on the 3rd Facing the Multicore Challenge conference