amrex
amrex copied to clipboard
GPU support for FillBoundary
In my code, FillBoundary takes most of the run time. I hope to transfer my code to GPU support. Now i am a little worried about the FillBoundary's efficiency. What i want to know is where the fillboundary will perfom. CPU or GPU?
Is it true that your code spends the majority of its time in FillBoundary even within a single node, or is that statement only true at large scale? If so, at what scale does it start to dominate?
In a single node, it usually takes above 35% time since i have optimize the other part.
you can also refer to #303 for more details.