Autotuning PolyBench benchmarks with LLVM Clang/Polly loop optimization pragmas using Bayesian optimization

We develop a ytopt autotuning framework that leverages Bayesian optimization to explore the parameter space search and compare four different supervised learning methods within evaluate their effectiveness. select six of most complex PolyBench benchmarks apply newly developed LLVM Clang/Polly loop pragmas optimize them. then use pragma parameters improve performance. The experimental results show our approach outperforms other compiling provide smallest execution time for syr2k, 3mm, heat-3d, lu, covariance with two large datasets in 200 code evaluations effectively searching spaces up 170,368 configurations. find Floyd–Warshall benchmark did not benefit from autotuning. To cope this issue, we some compiler option solutions Then present without user's knowledge using simple mctree further performance benchmark. also extend tune deep application.