Extending the Polyhedral Compilation Model for Debugging and Optimization of SPMD-style Explicitly-Parallel Programs

[摘要] The SPMD (Single Program Multiple Data) parallelism continues to be one of the most popular parallel execution models in use today, as exemplified by OpenMP for multi-core systems, CUDA and OpenCL for accelerator systems, and MPI for distributed-memory systems.The basic idea behind the SPMD model, which differentiates it from task-parallel models, is that all logical processors (worker threads) execute the same program with sequential code executed redundantly and parallel code executed cooperatively.This thesis extends the polyhedral compilation model to enable analysis of SPMD-style explicitly-parallel programs.This thesis demonstrates the value of this extended polyhedral model by describing its use in two applications to help developers of SPMD programs --- identification of data races, as well as identification and removal of redundant barriers.This thesis evaluates the effectiveness of these two applications using 34 OpenMP programs from the {t OmpSCR} and {t PolyBench/ACC} OpenMP benchmark suites.

[发布机构] Rice University

[效力级别] parallelism