Semi-Explicit Parallel Programming in a Purely Functional Style

Loidl, Hans-Wolfgang; Trinder, Phil; Kevin, Kevin; Zain, Abdallah Al; Baker-Finch, Clem

doi:10.1201/9781420064872.ch2

Cited by 1 publication

(1 citation statement)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MPI offers a large set of predefined communication patterns called collective operations. Many newer programming paradigms, such as PGAS languages [22,33,6,5], declarative parallel languages [17], or annotation-based languages [35,23] provide a conceptually simpler programming interface and rely on the compiler or the communication layer for optimizations. While some languages offer intrinsics or runtime libraries for several collective operations, many do not support such a specification at all moving the burden of pattern recognition to the compiler or forcing the user to hand-code non performance-portable algorithms.…”

Section: Introductionmentioning

confidence: 99%

Runtime detection and optimization of collective communication patterns

Hoefler

Schneider

2012

Proceedings of the 21st International Conference on Parallel Architectures and Compilation Techniques

View full text Add to dashboard Cite

Parallelism is steadily growing, remote-data access will soon dominate the execution time of large-scale applications. Many largescale communication patterns expose significant structure that can be used to schedule communications accordingly. In this work, we identify concurrent communication patterns and transform them to semantically equivalent but faster communications. We show a directed acyclic graph formulation for communication schedules and concisely define their synchronization and data movement semantics. Our dataflow solver computes an internal representation (IR) that is amenable to pattern detection. We demonstrate a detection algorithm for our IR that is guaranteed to detect communication kernels on subsets of the graph and replace the subgraph with hardware accelerated or hand-tuned kernels. Those techniques are implemented in an open-source detection and transformation framework to optimize communication patterns. Experiments show that our techniques can improve the performance of representative example codes by several orders of magnitude on two different systems. However, we also show that some collective detection problems on process subsets are NP-hard. The developed analysis techniques are a first important step towards automatic large-scale communication transformations. Our developed techniques open several avenues for additional transformation heuristics and analyses. We expect that such communication analyses and transformations will become as natural as pattern detection, just-in-time compiler optimizations, and autotuning are today for serial codes.

show abstract