OmniTune

Humans Need Not Apply*. Self-tuning programs are here, and they outperform human experts. OmniTune is a an extensible and distributed autotuner for runtime optimisation of parameters.

Papers

C. Cummins, P. Petoumenos, M. Steuwer, H. Leather. "Towards Collaborative Performance Tuning of Algorithmic Skeletons". HLPGPU, 2016.

C. Cummins, P. Petoumenos, M. Steuwer, H. Leather. "Autotuning OpenCL Workgroup Size for Stencil Patterns". ADAPT, 2016. arXiv.

C. Cummins. "Autotuning Stencil Codes with Algorithmic Skeletons". 2015.

Talks

C. Cummins. "Towards Collaborative Performance Tuning of Algorithmic Skeletons". HLPGPU, 2016.

C. Cummins. "Autotuning OpenCL Workgroup Size for Stencil Patterns". ADAPT, 2016.

Posters

C. Cummins, P. Petoumenos, M. Steuwer, H. Leather. "Autotuning OpenCL Workgroup Sizes". HiPEAC, 2016. Featured on Better Posters.

C. Cummins, P. Petoumenos, M. Steuwer, H. Leather. "Humans Need Not Apply". Google PhD Student Summit on Compiler & Programming Technology, 2015.

C. Cummins, H. Leather, P. Petoumenos, R. Mayr. "Can we achieve ease of use and high performance?". 2015.

AI researcher specializing in compilers and code optimization.