Advanced search
1 file | 48.76 KB

An experimental study on performance portability of OpenCL kernels

Sean Rul (UGent) , Hans Vandierendonck (UGent) , Joris D'Haene (UGent) and Koen De Bosschere (UGent)
Author
Organization
Abstract
Accelerator processors allow energy-efficient computation at high performance, especially for computationintensive applications. There exists a plethora of different accelerator architectures, such as GPUs and the Cell Broadband Engine. Each accelerator has its own programming language, but the recently introduced OpenCL language unifies accelerator programming languages. Hereby, OpenCL achieves functional protability, allowing to reduce the development time of kernels. Functional portability however has limited value without performance portability: the possibility to re-use optimized kernels with good performance. This paper investigates the specificity of code optimizations to accelerator architecture and the severity of lack of performance portability.
Keywords
openCL, multi-core, high-performance computing, gpu

Downloads

  • saahpc2010 final.pdf
    • full text
    • |
    • open access
    • |
    • PDF
    • |
    • 48.76 KB

Citation

Please use this url to cite or link to this publication:

Chicago
Rul, Sean, Hans Vandierendonck, Joris D’Haene, and Koen De Bosschere. 2010. “An Experimental Study on Performance Portability of OpenCL Kernels.” In Application Accelerators in High Performance Computing, 2010 Symposium, Papers.
APA
Rul, S., Vandierendonck, H., D’Haene, J., & De Bosschere, K. (2010). An experimental study on performance portability of OpenCL kernels. Application Accelerators in High Performance Computing, 2010 Symposium, Papers. Presented at the 2010 Symposium on Application Accelerators in High Performance Computing (SAAHPC  ’10).
Vancouver
1.
Rul S, Vandierendonck H, D’Haene J, De Bosschere K. An experimental study on performance portability of OpenCL kernels. Application Accelerators in High Performance Computing, 2010 Symposium, Papers. 2010.
MLA
Rul, Sean, Hans Vandierendonck, Joris D’Haene, et al. “An Experimental Study on Performance Portability of OpenCL Kernels.” Application Accelerators in High Performance Computing, 2010 Symposium, Papers. 2010. Print.
@inproceedings{1016024,
  abstract     = {Accelerator processors allow energy-efficient computation at high performance, especially for computationintensive applications. There exists a plethora of different accelerator architectures, such as GPUs and the Cell Broadband Engine. Each accelerator has its own programming language, but the recently introduced OpenCL language unifies accelerator programming languages. Hereby, OpenCL achieves functional protability, allowing to reduce the development time of kernels. Functional portability however has limited value without performance portability: the possibility to re-use optimized kernels with good performance. This paper investigates the specificity of code optimizations to accelerator architecture and the severity of lack of performance portability.},
  author       = {Rul, Sean and Vandierendonck, Hans and D'Haene, Joris and De Bosschere, Koen},
  booktitle    = {Application Accelerators in High Performance Computing, 2010 Symposium, Papers},
  keyword      = {openCL,multi-core,high-performance computing,gpu},
  language     = {eng},
  location     = {Knoxville, TN, USA},
  pages        = {3},
  title        = {An experimental study on performance portability of OpenCL kernels},
  url          = {http://saahpc.ncsa.illinois.edu/papers/paper\_2.pdf},
  year         = {2010},
}