Closed WeiqunZhang closed 1 month ago
I have tested with the latest CUDA, HIP and SYCL compilers. The compilers seem to have improved a lot. It used to be quite slow to compile the EB tensor and nodal projection solvers. But now I only saw a few percent difference when those solvers were not included.
We do not need to compile these linear solvers that are intended for incompressible flow solvers. This should speed up the build process a little bit.
Depends on https://github.com/AMReX-Codes/amrex/pull/4181