Section: Application Domains

Users of HPC Platforms and Scientific Computing

The convergence between embedded computing systems and high-performance computing (HPC) technologies offers new computing platforms and tools for the users of scientific computing (e.g., people working in numerical analysis, in simulation, modeling, etc.). The proliferation of “cheap” hardware accelerators and multicores makes the “small HPC” (as opposed to computing centers with more powerful computers, grid computing, and exascale computing) accessible to a larger number of users, even though it is still difficult to exploit, due to the complexity of parallel programming, code tuning, interaction with compilers, which result from the multiple levels of parallelism and of memories in the recent architectures. The link between compiler and code optimization research (as in Compsys) and such users are still to be reinforced, both to guarantee the relevance of compiler research efforts with respect to application needs, and to help users better interact with compiler choices and understand performance issues.

The support of Labex MILYON (through its thematic quarters, such as the thematic quarter on compilation we organized in 2013 (Thematic quarter on compilation: http://labexcompilation.ens-lyon.fr/), or the 2016 thematic quarter on high-performance computing, with a dedicated inter-disciplinary spring school between numerical simulation and polyhedral compilation, see hereafter) and the activities of the LyonCalcul initiative (Lyon Calcul federation: http://lyoncalcul.univ-lyon1.fr) are means to get closer to users of scientific computing, even if it is too early to know if Compsys will indeed be directly helpful to them.