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
some operations of indexation cannot be conducted because they break broadcast rule
readability is not great around mask and slicing boudaries
the slicing, though faster, leads to a more complex NPIR and codegen
debug-ability, on top of the issue of generated code, is not great
In order to create an area where quick iterative development of features can happen and start looking at debug tooling when writing model (which part of the tooling will need to address the nitty-gritty of the numerics) we propose to write a plain python "debug" backend.
This backend first and foremost shall be simplicity and closeness to C/Fortran looking code
The current
numpy
backend as limitations:mask
and slicing boudariesIn order to create an area where quick iterative development of features can happen and start looking at debug tooling when writing model (which part of the tooling will need to address the nitty-gritty of the numerics) we propose to write a plain python "debug" backend.
This backend first and foremost shall be simplicity and closeness to C/Fortran looking code
Parent: GEOS-ESM/SMT-Nebulae#36
gt:debug
backend straight from OIR or NPIRThe text was updated successfully, but these errors were encountered: