Input Sensitivity of GPU Program Optimizations
Computer Science Department, College of William and Mary
Cetus Users and Compiler Infastructure Workshop, PACT 2011
@article{liu2011input,
title={Input Sensitivity of GPU Program Optimizations},
author={Liu, Y. and Zhang, E.Z. and Bhamidipati, P. and Shen, X.},
booktitle={Cetus Users and Compiler Infastructure Workshop, PACT 2011},
year={2011}
}
Graphic Processing Units (GPU) have become increasingly adopted for the enhancement of computing throughput. However, the development of a high-quality GPU application is challenging, due to the large optimization space and complex unpredictable effects of optimizations on GPU program performance. Many recent efforts have been employing empirical search-based auto-tuners to tackle the problem, but few of them have concentrated on the influence of program inputs on the optimizations. In this paper, based on a set of CUDA and OpenCL kernels, we report some evidences on the importance for autotuners to adapt to program input changes, and present a framework, GADAPT+, to address the influence by constructing cross-input predictive models for automatically predicting the (near-)optimal configurations for an arbitrary input to a GPU program. G-ADAPT+ is based on source-to-source compilers, specifically, Cetus and ROSE. It supports the optimizations of both CUDA and OpenCL programs.
October 13, 2011 by hgpu
