Hybrid parallelization of In-Situ Adaptive Manifolds for computationally efficient turbulent combustion simulations

The cost of turbulent combustion simulations can be decreased by utilizing manifold-based combustion models, which project the thermochemical state onto a lower-dimensional manifold such that the thermochemical state can be computed separately from the flow solver. Traditionally, the manifold equation solutions for the thermochemical state were precomputed and pretabulated, resulting in large memory requirements and significant precomputation cost. In-Situ Adaptive Manifolds (ISAM) allows for solutions to the manifold equations to be computed as the flow simulation progresses and stored using In-Situ Adaptive Tabulation (ISAT), allowing for more general models to be used. The initial population of the ISAT databases can be accelerated by trading MPI processes for OpenMP threads due to faster computing of the solutions to the manifold equations and reduced redundancy in computing these solutions. However, once the ISAT databases are fully populated, the serialized ISAT algorithm leads to an overall increase in the computational cost of ISAM. In this work, a new adaptively threaded version of the ISAT algorithm has been developed to effectively leverage the OpenMP threads. Overall efficiency and scaling with the number of OpenMP threads will be demonstrated.