Second International Workshop on Reproducibility in Parallel Computing (REPPAR)

摘要

Conducting sound and reproducible experiments in parallel computing is not easy, as hardware and software architectures of current parallel computers are most often very complex. This high complexity makes it difficult-and often impossible-for computer scientists to model such systems mathematically. For that reason, scientists rely on experiments to study new parallel algorithms, different software solutions (e.g., operating systems), or novel hardware architectures. The situation in parallel computing is made even more difficult than it would be otherwise, as parallel systems are in a constant state of flux, e.g., the total core count is rapidly growing and many programming paradigms for parallel machines have emerged and are actively being used in a hybrid fashion, e.g., MPI, OpenMP, or PGAS.