Massive Computing Systems will be needed to maintain competitiveness in all areas of science, engineering or business. With top systems reaching the PFlop barrier, the next challenge is to understand how applications have to be implemented and be prepared for the ExaFlop target. Multicore chips are already here but will grow in the next decade to several hundreds of cores. Hundreds of thousands of nodes based on them will constitute the future exascale systems.
TEXT is centered on the vision that the key component to support high productivity and efficient use of a system is the programming model, and we defend that MPI/SMPSs is a hybrid approach that can today be demonstrated and show the way to follow on the path to exascale. The SMPSs model provides the necessary support for asynchrony and heterogeneity as well as enabling incremental parallelization, modularity and portability of applications. By integrating it within MPI we can propagate its characteristics to the global application level. This is also a way to leverage and provide a smooth migration path for the huge number of applications today written in MPI.
The focus of the TEXT project is to install the MPI/SMPSs environment at several HPC facilities of partners (also involved in the PRACE project) and demonstrate how seven real and relevant applications/ libraries can be improved using it. The codes fall in the areas of basic linear algebra libraries, geophysics, plasma physics, engineering and molecular dynamics. They have been selected considering the impact in their respective scientific communities.
We will actually validate our claim by evaluating the ported applications with different end users and collect their feedback to further improve the technology. We will also promote the use of the model to other application developers beyond what is feasible to integrate as committed project partners.
In order to exploit the opportunity window and really achieve global impact the project proposal is for 2 years. The TEXT proposal is composed of three networking activities, three service activities and two joint research activities.
Regarding networking activities, TEXT will focus on influencing programming model standards and promoting the technology demonstrated, while interacting with other international activities (i.e. IESP). TEXT will also perform training activities and organization of workshops to demonstrate to wider communities of application developers the many benefits of the proposed hybrid MPI/StarSs model.
Regarding service activities, TEXT considers three levels of developers/users: programming model/runtime developers, application developers and application users, with each level supporting the next level. Three service activities have been designed around these levels:
programming model infrastructure deployment
porting e_orts by application developers in the consortium
service/consulting to external users
Joint Research Activities
Finally, regarding joint research activities, TEXT will focus on improving the infrastructure (environment/platform/applications) initially available in the consortium. Two JRAs are proposed: one centered around programming development support and a second one centered on algorithmic optimization (i.e. use of asynchronous algorithms and support for load balancing).