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
Currently, 95% of the codebase is written in C++ and all of the "heavy lifting" is done by C++. The last step, however, is done in Python and requires a NumPy dependency. The code's entire Python dependency is found in the postprocess.py file in the root directory of the project.
It would be fantastic to convert this into C++ so that everything can be contained in a single compiled binary. This will include rewriting a bunch of simple NumPy functions in C++. There are also some subtleties, since for example, correct calculation of the ridge energy observables uses the np.ma.* modules (masked arrays) distinct from the standard np.*. These allow for masking (expected) nan values.
The other complication is that I want to keep external dependencies to nothing. The only libraries hdspin depends on externally are MPI. I really don't want the code to depend on any other external programs/libraries.
So, long story short:
Rewrite the postprocessing.py file in C++, integrating it into the hdspin source proper
Integrate the post processing operation into the main code so that it runs after the simulations are complete
[bonus] Make post processing MPI-enabled
Note: this is started here already: inc/array.h
The text was updated successfully, but these errors were encountered:
Currently, 95% of the codebase is written in C++ and all of the "heavy lifting" is done by C++. The last step, however, is done in Python and requires a NumPy dependency. The code's entire Python dependency is found in the
postprocess.py
file in the root directory of the project.It would be fantastic to convert this into C++ so that everything can be contained in a single compiled binary. This will include rewriting a bunch of simple NumPy functions in C++. There are also some subtleties, since for example, correct calculation of the ridge energy observables uses the
np.ma.*
modules (masked arrays) distinct from the standardnp.*
. These allow for masking (expected) nan values.The other complication is that I want to keep external dependencies to nothing. The only libraries hdspin depends on externally are MPI. I really don't want the code to depend on any other external programs/libraries.
So, long story short:
postprocessing.py
file in C++, integrating it into thehdspin
source properNote: this is started here already:
inc/array.h
The text was updated successfully, but these errors were encountered: