The --double-precision command line argument is not applied to the FFT
Closed this issue · 1 comments
mortele commented
The initialization of the pmesh
object should use something like dtype="f8" if dtype == np.float64 else "f4"
in
pm = pmesh.ParticleMesh(
config.mesh_size, BoxSize=config.box_size, dtype="f4", comm=comm
)
Currently it is just locked to using single precision floats always.