Abstract. Approaching the theoretical performance of hierarchical multicore machines requires a very careful distribution of threads and data among the underlying non-uniform architecture in order to minimize cache misses and NUMA penalties. While it is acknowledged that OpenMP can enhance the quality of thread scheduling on such architectures in a portable way, by transmitting precious information about the affinities between threads and data to the underlying runtime system, most OpenMP runtime systems are actually unable to efficiently support highly irregular, massively parallel applications on NUMA machines. In this paper, we present a thread scheduling policy suited to the execution of OpenMP programs featuring irregular and massive nested parallelism over hierarchical architectures. Our policy enforces a distribution of threads that maximizes the proximity of threads belonging to the same parallel section, and uses a NUMA-aware work stealing strategy when load balancing is needed. It has been developed as a plug-ins to the FORESTGOMP OpenMP platform [TBG + 07]. We demonstrate the efficiency of our approach with a highly irregular recursive OpenMP program resulting from the generic parallelization of a surface reconstruction application. We achieve a speedup of 14 on a 16-core machine with no application-level optimization.

Abstract. Thanks to recent advances in virtualization technologies, it is now possible to benefit from the flexibility brought by virtual machines at little cost in terms of CPU performance. However on HPC clusters some overheads remain which prevent widespread usage of virtualization. In this article, we tackle the issue of inter-VM MPI communications when VMs are located on the same physical machine. To achieve this we introduce a virtual device which provides a simple message passing API to the guest OS. This interface can then be used to implement an efficient MPI library for virtual machines. The use of a virtual device makes our solution easily portable across multiple guest operating systems since it only requires a small driver to be written for this device. We present an implementation based on Linux, the KVM hypervisor and Qemu as its userspace device emulator. Our implementation achieves near native performance in terms of MPI latency and bandwidth.

No abstract