GPU compatibility (questions, open branches, problems?)

[borjapedro]

Abstract and Motivation

As a brief introduction, my research group and myself (AIA, RWTH-Aachen University) are using Cantera for the computation of reaction rates and diffusion coefficients in our CFD solver. We have obtained great results using Cantera, however we will be porting our code to GPU in the near future (already working on it) in order to allow GPU and CPU alternatives. The code is written in C++, with MPI, Open-MP multithreading on CPU and PSTL based multithreading on GPU, focusing on performance for HPC computations. As far as I understand, this would be an issue with the current implementation of Cantera, which is based on CPU architecture and does not allow GPU acceleration.

More than a request, this is an open question to the Cantera community, with more knowledge in the matter: Is there any work being done as a group on a GPU porting of Cantera? Or does it already exist? We would gladly help in the development or testing if any such thing is being attempted.

I know this is probably not the best place for such a question, however could not think of an alternative to reach the Cantera community.

Thanks in advance for the help.