-
2.8477fccc1 · ·
add working framework for postprocessing Things are probably not fully stable in the backend, but I think we have a reasonable way to handle postprocessing. The important bit is that there is a transparent way to read either raw binary fields or HDF5, whatever is available for the given dataset, and then we can make children classes that will perform various tasks.
-
-
-
-
2.5.3db4271b5 · ·
Somewhat stable C++ classes Tagging after bugfixes and some changes made directly in develop because they were quite straightforward. At the moment, it looks like the `direct_numerical_simulation` class is quite mature, and we could, in principle, think about generating any sort of pseudo-spectral DNS as a child of it. I don't expect to make important changes to the prototype of `direct_numerical_simulation` in the future.
-
-
2.48305457c · ·
particle I/O is better handled now. The trick is to write using many processes, but not too many.
-
2.3be0115f6 · ·
we now have two versions of the DNS code both work with particles, but the new one has no sampling functionality yet
-
2.2da8dc217 · ·
New field-class based solver should now be preferred. I/O is done through checkpoints in HDF5, FFTW with threads is used, and all our tests say the new code scales quite nicely.
-
-
-
2.03846d148 · ·
Add new fluid solver. NSVorticityEquation uses HDF5 for files, and it uses the new field/kspace classes, which means we can separate backend development from other tasks more cleanly.