6633

Towards Automatic C Programs Optimization and Parallelization using the PIPS-PoCC Integration

Dounia Khaldi, Corinne Ancourt, Francois Irigoin
CRI, Mathematiques et Systemes, MINES ParisTech, 35 rue Saint-Honore, 77300. Fontainebleau, France
CRI, Mathematiques et Systemes, MINES ParisTech, report A/448/CRI, 2011

@article{khaldi2011towards,

   title={Towards Automatic C Programs Optimization and Parallelization using the PIPS-PoCC Integration},

   author={Khaldi, D. and Ancourt, C. and Irigoin, F.},

   year={2011}

}

This paper explains how the PIPS source-to-source compilation framework integrates the Polyhedral Compiler Collection (PoCC) as one of PIPS many program transformations. The integration between PIPS and PoCC extracts automatically the static control parts of the source code, which can be optimized independently by PoCC and then reintegrates them transparently in the user source code. PIPS can either be used simply as a wrapper around PoCC to simplify the compilation process, or use PoCC as a step in more complex compilation schemes such as heterogeneous code generation for CUDA or FPGA based-machine. This is possible when the polyhedral model can be used to fine-tune fine-grain parallelism and memory locality. This paper explores the issues in the design of the integration and presents results for some benchmarks of the OpenGPU project.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: