Skip to content

feat(mtagro): remove vesin#3

Merged
HaoZeke merged 17 commits intometatomicfrom
noVesin
Feb 3, 2026
Merged

feat(mtagro): remove vesin#3
HaoZeke merged 17 commits intometatomicfrom
noVesin

Conversation

@HaoZeke
Copy link
Member

@HaoZeke HaoZeke commented Jan 30, 2026

No vesin involved, so should be good for upstream too. Also got faster... and takes the cutoff from the model, so win-win.

Core t (s)   Wall t (s)        (%)
       Time:        4.751        0.297     1599.8
                 (ns/day)    (hour/ns)    (ms/step)  (Matom*steps/s) 
Performance:       29.382        0.817        2.941            0.043

So 2x the first vesin version.

Same design as #1.

For instructions on optimal usage / reproduction consider details here.

@HaoZeke HaoZeke changed the base branch from metatomic to useVesin January 30, 2026 11:43
Copy link
Contributor

@PicoCentauri PicoCentauri left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks clean!

Base automatically changed from useVesin to metatomic January 30, 2026 18:43
@HaoZeke HaoZeke requested a review from Luthaf February 2, 2026 14:58
real* data_ptr = reinterpret_cast<real*>(positions_.data());
const size_t n_reals = static_cast<size_t>(n_atoms) * 3;
mpiComm_.sumReduce(gmx::ArrayRef<real>(data_ptr, data_ptr + n_reals));
mpiComm_.sumReduce(3 * numInput, positions_.data()->as_vec());
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So sumReduce is used to do a gather here, correct? There is no gather primitive instead? (I guess it does not matter since this will go away)

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not from a quick reading but also yeah it's going the way of the dodo anyway

@HaoZeke HaoZeke requested a review from Luthaf February 2, 2026 16:26
Copy link
Member

@Luthaf Luthaf left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is in a good state, let's merge and keep improving it in the next PR!

@HaoZeke HaoZeke merged commit 4d8b50d into metatomic Feb 3, 2026
@HaoZeke HaoZeke deleted the noVesin branch February 3, 2026 14:40
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.

3 participants