Paris March 18, 2011 – HPC Project introduces the 1.1 version of Par4All open parallelization platform with automatic code generation for the NVIDIA GPU based hybrid computing architectures. This version brings a significant performance improvement on the execution time of the generated code.
Par4All is the industrial implementation of a technology developed by the teams of CRI/Mines ParisTech and Télécom Bretagne. Its goal is to allow industrial users to meet the challenge of multi-cores architectures and other parallel processors for generic or embedded systems. From a C or FORTRAN application, Par4All automatically generates a parallel code to OpenMP or CUDA (compilable on NVIDIA GPU). The generated code is readable and completely traceable with the original code. The whole process works like a usual compilation.
The code generated with Par4All version 1.1 is significantly faster at execution time. This is due to the new static data-flow analyzer. This new analyzer allows the optimization of the communications between host and GPU as well as allocations to the GPU. Practical examples of applications are available on http://www.par4all.org/documentation/benchmarks/
HPC Project was established in December 2007. HPC Project is a pioneer in developing tools and strategies for high performance computing and code optimization. HPC Project goal is to bring the power of supercomputer on the engineer’s desk.
roger.marhuenda (at) hpc-project (dot) com