The open archive for STFC research publications

Full Record Details

DOI 10.5286/raltr.2021001
Persistent URL http://purl.org/net/epubs/work/49338008
Record Status Checked
Record Id 49338008
Title A comparison of frameworks for hybrid programming with Graphics Processing Units
Abstract Over the years, Graphics Processing Unit (GPU) technology has become increasingly complex, attaining significant speedup against traditional multicore CPUs. General purpose GPUs (GPGPUs) leverage streaming multiprocessors with their own on-chip shared memory to perform large scale multi-threaded operations. However, writing code to explicitly make use of GPU architectures is often challenging and requires advanced knowledge in GPGPU programming. In this study, we will outline the various parallel programming frameworks available for targeting GPU resources and compare both the computational performance and the productivity cost required to develop the accelerator code. The aim is to provide a comprehensive resource for members of the scientific computing community, serving as a guide in identifying the best approach to deploy when porting scientific codes; in terms of productivity and performance.
Organisation STFC , HC
Funding Information
Related Research Object(s):
Licence Information: Creative Commons Attribution 4.0 International (CC BY 4.0)
Language English (EN)
Type Details URI(s) Local file(s) Year
Report RAL Technical Reports RAL-TR-2021-001. STFC, 2021. RAL-TR-2021-001.pdf 2021