Skip to content

Refactor: Use psi_initializer instead of wavefunc #5775

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 28 commits into from
Jan 9, 2025
Merged

Refactor: Use psi_initializer instead of wavefunc #5775

merged 28 commits into from
Jan 9, 2025

Conversation

Qianruipku
Copy link
Collaborator

@Qianruipku Qianruipku commented Dec 27, 2024 •
edited
Loading

Refactor psi_initializer
Remove wavefunc and wf_atomic
Fix #5811
Fix #5578

@kirk0830
Copy link
Collaborator

@Qianruipku Hi, qianrui. I am happy to see you are trying to use psi_initializer in SDFT. Here I share some information and my findings in previous days when I was trying to make it available on GPU.

  1. There is a bug/ a hidden undefined behavior in function random_t that will be triggered in case nspin 4. The energy will vary in a random manner if you run with MPI. To detect this, you can run the integrated test case 140_PW_15_SO.
  2. Because I use smart pointer to manage the projected NAO in psi_initializer to ensure the large piece of memory is safely released when program runs out of life span of psi_initializer instance, it is not directly feasible to make it run on GPU, and I am considering roll back to use raw pointer (although regret to say it is a unsafe choice...), but not manage to find enough time to do this.

Regards.

@Qianruipku Qianruipku changed the title Try using psi_initializer Refactor: Use psi_initializer instead of wavefunc Jan 4, 2025
@Qianruipku Qianruipku marked this pull request as ready for review January 4, 2025 14:28
@mohanchen mohanchen added Features Needed The features are indeed needed, and developers should have sophisticated knowledge Refactor Refactor ABACUS codes The Absolute Zero Reduce the "entropy" of the code to 0 labels Jan 6, 2025
@mohanchen mohanchen merged commit 0401194 into deepmodeling:develop Jan 9, 2025
14 checks passed
@Cstandardlib Cstandardlib mentioned this pull request Jan 9, 2025
Merged
@Qianruipku Qianruipku mentioned this pull request Jan 10, 2025
Closed
9 tasks
Fisherd99 pushed a commit to Fisherd99/abacus-BSE that referenced this pull request Mar 31, 2025
@Qianruipku
Refactor: Use psi_initializer instead of wavefunc (deepmodeling#5775)
b0a7450 
* use psi_initializer

* fix compile

* same results of random init

* make atomic initialized results right

* finish refactor

* fix compile

* fix compile

* fix UTs

* update results

* update results

* update GPU results

* update

* refactor pw

* change 108_PW_RE_PINT_RKS results

* update results

* remove openmp for random generate

* update

* remove psi_initializer in Doc

* remove omp2

* fix compile
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@mohanchen mohanchen mohanchen approved these changes

Assignees
No one assigned
Labels
Features Needed The features are indeed needed, and developers should have sophisticated knowledge Refactor Refactor ABACUS codes The Absolute Zero Reduce the "entropy" of the code to 0
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Current PW version does not support single precision Bug: init_wfc nao doesn't support gpu
3 participants
@Qianruipku @kirk0830 @mohanchen