Helium-Solute NNP for interaction [cschran/lduran/hforbert] #3043
Conversation
| TYPE(mp_comm_type) :: new_comm | ||
| TYPE(section_vals_type), POINTER :: helium_section, input_worm | ||
| TYPE(cp_logger_type), POINTER :: logger, tmplogger | ||
| TYPE(mp_para_env_type), POINTER :: new_comm |
There was a problem hiding this comment.
Why do you make this communicator a pointer? In your case you know that all pointers point to the same object. The others are just mere copies. I also do not see when you deallocate the pointer.
There was a problem hiding this comment.
I make that a pointer, because cp_logger_create needs a pointer and I cannot make it a target since it's part of a structure which I didn't want to make a target. I didn't deallocate because I wrongly assumed %free is enough :)
But I agree the pointers are a mess here and also the parallization (helium assumes only one rank per environment
and doesn't use mpi, which clashes with the regular nnp (hence I had to create a logger with a single rank, so nnp is happy).
There was a problem hiding this comment.
Ah, thanks. No, %free just frees the data of the object but it cannot free the pointer (or the object itself). And I mixed up mp_para_env_type and mp_comm_type. Then, it is ok by design.
Helium-Solute NNP for interaction [cschran/lduran/hforbert]
3 participants
No description provided.