Skip to content

[pull] develop from lammps:develop#122

Merged
pull[bot] merged 55 commits intocomphy-lab:developfrom
lammps:develop
Mar 30, 2026
Merged

[pull] develop from lammps:develop#122
pull[bot] merged 55 commits intocomphy-lab:developfrom
lammps:develop

Conversation

@pull
Copy link
Copy Markdown

@pull pull bot commented Mar 30, 2026

See Commits and Changes for more details.


Created by pull[bot] (v2.0.0-alpha.4)

Can you help keep this open source service alive? 💖 Please sponsor : )

stanmoore1 and others added 30 commits February 26, 2026 09:50
this adds a new bit flag ENERGY_ONLY to the eflag variable
and adds code to force computing styles to convert this into
an eflag_only variable that is zero or non-zero based on eflag
for fixes and computes calling force styles' compute() method
and only use the global energy add ENERGY_ONLY bit to eflag.
Fixing error of model living on the CPU side rather than GPU side, which triggered

RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:0! (when checking argument for argument mat1 in method wrapper_CUDA_addmm)
Add supernormal() helper to compute correct surface normals for
superellipsoids using the gradient of the implicit surface function:
((|x/a|^e + |y/b|^e)^(n/e) + |z/c|^n)^(1/n) = 1

Previously the draw method used ellipsoid normals (gradient of
x^2/a^2 + y^2/b^2 + z^2/c^2) even when blockiness parameters were
provided for superellipsoids. The new supernormal() function correctly
computes the gradient components with the proper exponents.

Also refactored the draw loop to eliminate duplicated vertex scaling
and rotation/translation code between the dotri and doframe blocks.

Co-authored-by: akohlmey <289251+akohlmey@users.noreply.github.com>
Agent-Logs-Url: https://github.com/akohlmey/lammps/sessions/fcc6704a-49d1-43b7-94a8-40160dfe181d
akohlmey and others added 25 commits March 27, 2026 17:49
this is achieved by rotating the triangulated sphere topology so
that the three principal axes pass through the center or near the
center of a triangle. This makes the approximations look less
"pointy" especially for ellipsoids with prolate shapes
Create dihedral_nharmonic_kokkos.h and .cpp in src/KOKKOS/ following
the pattern from dihedral_multi_harmonic_kokkos and dihedral_fourier_kokkos.
Uses 2D Kokkos views for variable-length coefficient arrays and 1D
int views for per-type term counts. Update documentation to list the
new nharmonic/kk accelerated variant.

Agent-Logs-Url: https://github.com/akohlmey/lammps/sessions/5d795956-0b37-4808-837f-9a840867ed96

Co-authored-by: akohlmey <289251+akohlmey@users.noreply.github.com>
Add ENERGY_ONLY bit to eflag parameter for force computation
Add KOKKOS version of dihedral style nharmonic
Allow KOKKOS computes to pack/unpack forward comm buffers on device
Fix mliap/pytorch.py and move model to the GPU device
@pull pull bot locked and limited conversation to collaborators Mar 30, 2026
@pull pull bot added the ⤵️ pull label Mar 30, 2026
@pull pull bot merged commit 2bf8c4e into comphy-lab:develop Mar 30, 2026
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants