You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In the master branch (2decomp library and pencils domain decomposition), vertically summed slices Zsummed are transposed compared to other slices in the XY plane, ZSlice. For Zsummed slices, the fast index in memory corresponds to x (and y is slow).
This does not happen with the WiP branch (fftw-mpi and slabs domain decomposition), where the output has a consistent memory layout: x is slowest, then y, then z is fastest.
The text was updated successfully, but these errors were encountered:
In the master branch (
2decomp
library and pencils domain decomposition), vertically summed slicesZsummed
are transposed compared to other slices in the XY plane,ZSlice
. ForZsummed
slices, the fast index in memory corresponds tox
(andy
is slow).This does not happen with the WiP branch (
fftw-mpi
and slabs domain decomposition), where the output has a consistent memory layout:x
is slowest, theny
, thenz
is fastest.The text was updated successfully, but these errors were encountered: