Snowpack: efficient parameter choice for GPU kernels via static analysis and statistical prediction

Ranvijay Singh, Paul Wood, Ravi Gupta, Saurabh Bagchi, Ignacio Laguna

Abstract

The running time of GPU kernels depends on an invocation parameter, the number of threads in each thread block. Sometime the dependence is quite strong leading to 50--100% change in execution time for long-running kernels. Till now, it has been an art form to decide on the optimal setting for this parameter. NVIDIA provides a tool for CUDA kernels, called OCC, that guides a developer toward this goal. In this paper, we show that OCC maximizes occupancy of GPU cores but does not meet the performance goal in a wide class of applications. We develop a solution called Snowpack that uses static features in a statistical learning framework to choose the optimal block size parameter. It does this without needing to execute the kernel multiple times, as a possible alternate solution Autotuner does. We evaluate our solution, Snowpack on 89 kernels of 10 applications.
BibTeX
@inproceedings{singh2017snowpack,
  author = {Ranvijay Singh and Paul Wood and Ravi Gupta and Saurabh Bagchi and Ignacio Laguna},
  title = {{Snowpack: efficient parameter choice for GPU kernels via static analysis and statistical prediction}},
  booktitle = {ScalA@SC},
  year = {2017},
  doi = {10.1145/3148226.3148235},
}

Loading related content...

Figure