Section: New Results

Programming Many-core Systems

Participants : Jean-François Méhaut, Vania Marangozova-Martin.

Memory Affinity for Hierarchical Shared Memory Multiprocessors

Currently, parallel platforms based on large scale hierarchical shared memory multiprocessors with Non-Uniform Memory Access (NUMA) are becoming a trend in scientific High Performance Computing (HPC). Due to their memory access constraints, these platforms require a very careful data distribution. Many solutions were proposed to resolve this issue. However, most of these solutions did not include optimizations for numerical scientific data (array data structures) and portability issues. Besides, these solutions provide a restrict set of memory policies to deal with data placement. In [39] , we describe an user-level interface named Memory Affinity interface (MAi), which allows memory affinity control on Linux based cache-coherent NUMA (ccNUMA) platforms. Its main goals are, fine data control, flexibility and portability. The performance of MAi is evaluated on three ccNUMA platforms using numerical scientific HPC applications, the NAS Parallel Benchmarks and a Geophysics application. The results show important gains (up to 31%) when compared to the Linux default solution.

In [27] , we apply this memory affinity interface to an application that simulations seismic wave propagation. First, we parallelize the application using OpenMP. Then, we improve the OpenMP solution using the MAI (Memory Affinity Interface) library, which allows a control of memory allocation in NUMA machines. The results show that the optimization of memory allocation leads to significant performance gains over the pure OpenMP parallel solution.

Component-based Observations of Systems-on-a-Chip

We propose in [44] a component-based approach to provide a well-suited solution to the programming and deployment problems of systems on chip (SoC) that can become increasingly complex and heterogeneous. Focusing on the aspect of observation, we show, from system to application, that components help in observing all software levels. We present the EMBera prototype and relate our experience in implementing it on two different platforms: a Linux-based 16-core SMP machine and a 5-core embedded system developed by STMicroelectronics.