AUMA is a machine learning based auto tuner, which can be used to tune the performance of arbitrary applications.
How AUMA works is described in the following papers:
- Falch T. L., Elster A. C. Machine Learning Based Autotuning for OpenCL Performance Portability, IPDPSWS 15
- Falch T. L., Elster A. C. Machine learning based auto-tuning for enhanced performance portability of OpenCL applications, Concurrency and Computation, Practice and Experience, 2016
More detailed documentation on how to install and run AUMA and the included benchmarks can be found in the doc subdirectory.
The doc subdirectory also contains a report with additional data for the second paper above.
AUMA was written by Thomas L. Falch at the Norwegian University of Science and Technology. For conditions of distribution and use, see the accompanying LICENSE file.