Release of StarPU 1.0rc1
The first release candidate of the upcoming StarPU 1.0.
CAPS supports OpenACC
CAPS entreprise is part of the consortium promoting the OpenACC programming Standard together with NVidia, Cray and PGI.
Seventh ProHMPT meeting
ProHMPT partners gathering scheduled in Bordeaux on 27th and 28th of October, 2011.
BigDFT version 1.6 released
The 1.6 version of the BigDFT parallel electronic structure code has been released for the CECAM tutorial.
Sixth ProHMPT meeting
ProHMPT partners gathering scheduled at Montbonnot on 7th of July, 2011.
OpenHMPP open standard
CAPS entreprise announce the OpenHMPP open standard initiative.
Release of StarPU 0.9.1
The 0.9.1 release of StarPU provides a reduction mode, an external API for schedulers, theoretical bounds, power-based optimization, parallel tasks, an MPI DSM, profiling interfaces, an initial support for CUDA4 (GPU-GPU transfers), improved documentation and of course various fixes.
BigDFT version 1.5.2 released
The 1.5.2 version of the BigDFT parallel electronic structure code has been released.
Fifth ProHMPT meeting
ProHMPT partners gathering scheduled at CESTA/Le Barp on 14th of December and at INRIA Bordeaux on 15th of December, 2010.
Presentation of MAQAO at SuperComputing
The MAQAO analysis toolkit has been presented at SuperComputing.
Fourth ProHMPT meeting
ProHMPT partners gathering scheduled in Rennes at CAPS entreprise, on 29th of September, 2010.
ViTE (Visual Trace Explorer) 1.2 releases
The 1.2 release of the Visual Trace Explorer (ViTE).
Release of StarPU 0.4
The 0.4 release of StarPU provides support for task-based dependencies, implicit data-based dependencies (RAW/WAR/WAW), profiling feedback, an MPI layer, OpenCL and Windows support, as well as an API naming revamp.
SGPU 1 released
The first version of the SGPU support has been released.
BigDFT version 1.4 released
The 1.4 version of the BigDFT parallel electronic structure code has been released with OpenCL support.
Third ProHMPT meeting
ProHMPT partners gathering scheduled in Bordeaux on 1st and 2nd of June, 2010.
hwloc 1.0 released
The 1.0 version of the hwloc library for platform topology discovery has been released.
Computation Hours Grant
GENCI granted computation hours to the ProHMPT project for the Year 2010 on the Titane machine at CCRT.
Second ProHMPT meeting
ProHMPT partners gathering scheduled on 5th and 6th of October, 2009.
StarPU Team enters the NVIDIA Professor Partnership Program
The StarPU Team application for the NVIDIA Professor Partnership Program has been successful!
This partnership will help in improving CUDA support inside StarPU and in contributing research and experiments feedback to the CUDA development team. Moreover, NVidia will donate several high-end CUDA-capable cards!
ViTE (Visual Trace Explorer) 1.0 releases
ForestGOMP 0.1 and Marcel 2.90 releases
The first release of the ForestGOMP OpenMP runtime for the GNU OpenMP compiler is out, along with the 2.90 version of its companion multithreading library Marcel.
Release of StarPU 0.1
The first release of the StarPU unified runtime system for heterogeneous multicore architectures is out.
Project ProHMPT started
Project ProHMPT officially started on January 1st, 2009.
March meeting talks online
ProHMPT partners gathering on 12th and 13th of March, 2009.
First ProHMPT meeting
ProHMPT partners first gathering scheduled on 12th and 13th of March, 2009.
Programming Heterogeneous MultiProcessor Technologies.
The nature of computing architectures is currently enduring a deep transformation and an increase in creativity that radically breaks and ends, at least for a foreseeable future, the long era of stability and homogeneity that we have known. Multiprocessing inside a computing node is generalising, densifying and diversifying. The combined action of these three evolutions show the specificity and strength of the break and call for a necessary redesign of application programming methods, for the proposal of new programming models and more generally for an ambitious answer in phase with the potential of computing power enclosed in those new heterogeneous multicore technologies.
The ProHMPT project integrates itself in this effort from the community. It ambitions to take part as a major actor, thanks to a strong collaboration with CEA INAC (Grenoble, France) build on the momentum generated by the nanotechnologies. The extreme, and real, needs in terms of computing power expressed by the nanosimulations, at the very beginning from the numerous end-application fields of the nanotechnologies, imply to go beyond classical parallel programming and explore the exploitation of heterogeneous multicores such as the IBM Cell and the specialized processors such as the Graphical Processing Units (GPU).
Those architectures impose strong and complex constraints to the programmer, however, and existing tools to program such architectures are still rather limited. ProHMPT henceforth proposes to have French experts teams at such strategic levels as are compilers, runtime supports, and scientific computing kernels meet together with experts from nanosimulation, and work on the proposal of new approaches, methodologies, and tools to bring the expected abstraction and performance portability. The generality of the results obtained, beyond nanosimulation, will be validated in cooperation with CEA CESTA (Le Barp, France), which is greatly interested by the perspective of integrating and valorising the work of ProHMPT on an aerodynamic simulation code. The Bull company will contribute its valuable knowledge of hardware in the context of HPC, and its capacity of commercial diffusion to help make this project a success.