Closed pdziekan closed 1 year ago
In 3D velocities are copied from kij to kji storage order before saving and they are copied from kji to kij to reading. That's a lot of overhead that could be avoided by some smart hyperslab when reading in 3D with MPI
In 3D velocities are copied from kij to kji storage order before saving and they are copied from kji to kij to reading. That's a lot of overhead that could be avoided by some smart hyperslab when reading in 3D with MPI