Massively parallel simulations of binary black holes with Dendro-GR

Dendro-GR uses Wavelet Adaptive Multi-Resolution (WAMR) to generate an unstructured grid adapted to the spacetime geometry together with an octree based data structure. In this talk, we describe the current Dendro-GR algorithm, as well as its advantages and challenges when compared to traditional box-in-box AMR for simulating high-mass ratio black-hole binaries. The advantages include a significantly improved scaling and an associated reductions in "time to solution", as well as a more efficient use of computational resources. Among the challenges (which are not unique to Dendro-GR) is finding an automatic refinement function that efficiently produces high-resolution grids in the required regions without over resolving other regions.