Extending OpenMP for NUMA Machines

Extending the EPCC OpenMP Microbenchmarks for OpenMP 3.0

OpenMP task scheduling strategies for multicore NUMA systems

The recent addition of task parallelism to the OpenMP shared memory API allows programmers to express concurrency at a high level of abstraction and places the burden of scheduling parallel execution on the OpenMP run time system. Efficient scheduling of tasks on modern multi-socket multicore shared memory systems requires careful consideration of an increasingly complex memory hierarchy, inclu...

Volume Driven Data Distribution for NUMA-Machines

Highly scalable parallel computers, e.g. SCI-coupled workstation clusters, are NUMA architectures. Thus good static locality is essential for high performance and scalability of parallel programs on these machines. This paper describes novel techniques to optimize static locality at compilation time by application of data transformations and data distributions. The metric which guides the optim...

Evaluation of OpenMP Task Scheduling Algorithms for Large NUMA Architectures

Current generation of high performance computing platforms tends to hold a large number of cores. Therefore applications have to expose a fine-grain parallelism to be more efficient. Since version 3.0, the OpenMP standard proposes a way to express such parallelism through tasks. Because the task scheduling strategy is implementation defined, each runtime can have a different behavior and effici...

Hierarchical loop scheduling for clustered NUMA machines

Loop scheduling is an important issue in the development of high performance multiprocessors. As modern multiprocessors have high and non-uniform memory access (NUMA) costs, the communication costs dominate the execution of parallel programs. Previous anity algorithms perform better than dynamic algorithms under non-clustered NUMA multiprocessors, but they su€er heavy overheads when migrating ...