Projects per year
Abstract / Description of output
Selecting an appropriate workgroup size is critical for the performance of OpenCL kernels, and requires knowledge of the underlying hardware, the data being operated on, and the implementation of the kernel. This makes portable performance of OpenCL programs a challenging goal, since simple heuristics and statically chosen values fail to exploit the available performance. To address this, we propose the use of machine learning-enabled autotuning to automatically predict workgroup sizes for stencil patterns on CPUs and multi-GPUs.
We present three methodologies for predicting workgroup sizes. The first, using classifiers to select the optimal workgroup size. The second and third proposed methodologies employ the novel use of regressors for performing classification by predicting the runtime of kernels and the relative performance of different workgroup sizes, respectively. We evaluate the effectiveness of each technique in an empirical study of 429 combinations of architecture, kernel, and dataset, comparing an average of 629 different workgroup sizes for each. We find that autotuning provides a median 3.79× speedup over the best possible fixed workgroup size, achieving 94% of the maximum performance.
We present three methodologies for predicting workgroup sizes. The first, using classifiers to select the optimal workgroup size. The second and third proposed methodologies employ the novel use of regressors for performing classification by predicting the runtime of kernels and the relative performance of different workgroup sizes, respectively. We evaluate the effectiveness of each technique in an empirical study of 429 combinations of architecture, kernel, and dataset, comparing an average of 629 different workgroup sizes for each. We find that autotuning provides a median 3.79× speedup over the best possible fixed workgroup size, achieving 94% of the maximum performance.
Original language | English |
---|---|
Title of host publication | International Workshop on Adaptive Self-tuning Computing Systems (ADAPT) 2016 @ HiPEAC, Prague, Czech Republic, January 18, 2016 |
Number of pages | 8 |
Publication status | Published - 18 Jan 2016 |
Event | 6th International Workshop on Adaptive Self-tuning Computing Systems 2016 - Prague, Czech Republic Duration: 18 Jan 2016 → 18 Jan 2016 http://adapt-workshop.org/index2016.html |
Conference
Conference | 6th International Workshop on Adaptive Self-tuning Computing Systems 2016 |
---|---|
Abbreviated title | ADAPT 2016 |
Country/Territory | Czech Republic |
City | Prague |
Period | 18/01/16 → 18/01/16 |
Internet address |
Fingerprint
Dive into the research topics of 'Autotuning OpenCL Workgroup Size for Stencil Patterns'. Together they form a unique fingerprint.Projects
- 2 Finished
-
Distributed Heterogeneous Vertically IntegrateD ENergy Efficient Data centres
O'Boyle, M., Grot, B., Leather, H. & Viglas, S.
31/12/14 → 30/12/16
Project: Research
-