Efficient Execution of OpenMP on GPUs

OpenMP is the preferred choice for CPU parallelism in High-Performance-Computing (HPC) applications written in C, C++, or Fortran. As HPC systems became heterogeneous, OpenMP introduced support for accelerator offloading via the target directive. This allowed porting existing (CPU) code onto GPUs, i...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2022 IEEE/ACM International Symposium on Code Generation and Optimization (CGO) S. 41 - 52
Hauptverfasser: Huber, Joseph, Cornelius, Melanie, Georgakoudis, Giorgis, Tian, Shilei, Diaz, Jose M Monsalve, Dinel, Kuter, Chapman, Barbara, Doerfert, Johannes
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 02.04.2022
Schlagworte:
Analytical models
Codes
GPU
Graphics processing units
LLVM
Offloading
OpenMP
Optimization
Parallel processing
Runtime
Semantics
Static analysis
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:OpenMP is the preferred choice for CPU parallelism in High-Performance-Computing (HPC) applications written in C, C++, or Fortran. As HPC systems became heterogeneous, OpenMP introduced support for accelerator offloading via the target directive. This allowed porting existing (CPU) code onto GPUs, including well established CPU parallelism paradigms. However, there are architectural differences between CPU and GPU execution which make common patterns, like forking and joining threads, single threaded execution, or sharing of local (stack) variables, in general costly on the latter. So far it was left to the user to identify and avoid non-efficient code patterns, most commonly by writing their OpenMP offloading codes in a kernel-language style which resembles CUDA more than it does traditional OpenMP.In this work we present OpenMP-aware program analyses and optimizations that allow efficient execution of the generic, CPU-centric parallelism model provided by OpenMP on GPUs. Our implementation in LLVM/Clang maps various common OpenMP patterns found in real world applications efficiently to the GPU. As static analysis is inherently limited we provide actionable and informative feedback to the user about the performed and missed optimizations, together with ways for the user to annotate the program for better results. Our extensive evaluation using several HPC proxy applications shows significantly improved GPU kernel times and reduction in resources requirements, such as GPU registers.
DOI:10.1109/CGO53902.2022.9741290