Skip to content

Comments

Ability to change the total number of orbitals.#1

Open
danielhollas wants to merge 11 commits intocxlsmiles:masterfrom
PHOTOX:master
Open

Ability to change the total number of orbitals.#1
danielhollas wants to merge 11 commits intocxlsmiles:masterfrom
PHOTOX:master

Conversation

@danielhollas
Copy link

Hi @cxlsmiles!

I've made a couple small changes in your code. If you are interested you can merge them to your repo (I think should be pretty easy just by pushing a button).

The biggest change: ability to manually set the total number of MOs via n_mo parameter in input.txt
This is super useful for larger systems such as water pentamer where the orbitals cannot fit to RAM.
To make this work, I modified the subroutine that reads the MOs so that it automatically skips the orbitals with the index out of bounds (which would otherwise crash the code). This should be a backwards-compatible change but let me know if you have concerns.

I tried to make commits pretty small but let me know if some change is unclear.

I hope you're doing well! 🙂

It is not used anyway as per commit 29856a0.
 + add possibility to set n_mo manually in input.txt
(control section)

This is usefull for bigger systems for which the number of ints
is too big to fit in RAM.
It would be good to reduce RAM usage in the future,
there is definitely room for improvement.
+ use globals more wisely in read_hf_data module
Program would segfault due to incorrect array indexing
This is critical for larger systems
that will otherwise consume huge chunks of RAM
for 2e integrals.

We could do even better cause we do not need most
of the 2e integrals so we could store the sparse 2e matrix
in a more clever way, for example by reindexing virtual orbitals
and working only with (n_ov+1) x (n_ov+1) matrix for each
diagonalization step.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant