Skip to content

Refactor: Optimize memory management in RT-TDDFT to enable large-scale GPU calculations#6995

Merged
mohanchen merged 2 commits intodeepmodeling:developfrom
AsTonyshment:td_for_polaris
Mar 6, 2026
Merged

Refactor: Optimize memory management in RT-TDDFT to enable large-scale GPU calculations#6995
mohanchen merged 2 commits intodeepmodeling:developfrom
AsTonyshment:td_for_polaris

Conversation

@AsTonyshment
Copy link
Collaborator

@AsTonyshment AsTonyshment commented Mar 5, 2026

What's changed?

  • This PR addresses frequent Out-of-Memory (OOM) errors encountered during GPU-accelerated RT-TDDFT calculations for large-scale systems (e.g., >1000 Si atoms). The previous implementation suffered from a high peak memory footprint caused by redundant allocations of full-sized matrices and long-lived temporary tensors.

@mohanchen mohanchen added Refactor Refactor ABACUS codes GPU & DCU & HPC GPU and DCU and HPC related any issues Memory Memory issues labels Mar 6, 2026
@mohanchen
Copy link
Collaborator

mohanchen commented Mar 6, 2026

LGTM

@mohanchen mohanchen merged commit 7035e85 into deepmodeling:develop Mar 6, 2026
15 checks passed
@AsTonyshment AsTonyshment deleted the td_for_polaris branch March 6, 2026 10:06
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

GPU & DCU & HPC GPU and DCU and HPC related any issues Memory Memory issues Refactor Refactor ABACUS codes

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@AsTonyshment @mohanchen