Skip to content

v2026.6 — MatPL NEP/KK Release

Choose a tag to compare

View all tags
@pfsuo pfsuo released this 30 Jun 04:06
· 57 commits to main since this release
MatPL-pro-v2026.6
33797fc

MatPL NEP/KK v2026.6

Release of the NEP/KK LAMMPS Kokkos interface. The Pro version exhibits a 2‑ to 3‑fold increase in speed relative to the standard version.
matpl_vs_pro_all_in_one_giant_bars
speedup_ratio_heatmap

Pre-built libnep_gpu.so

Two pre-compiled variants for different GPU architectures:

File CUDA SM Targets Supported GPUs
cuda-11.8/libnep_gpu.so 11.8 35,37,60,61,70,75,80,86,89,90 K40/K80, P100, V100, GTX 1080, TITAN V, T4, RTX 2080, A100, A6000, RTX 3090, RTX 4090, H100
cuda-12.8/libnep_gpu.so 12.8 60,61,70,75,80,86,89,90,100,101,120 P100, V100, GTX 1080, TITAN V, T4, RTX 2080, A100, A6000, RTX 3090, RTX 4090, H100, B100/B200 (sm_100), RTX 5080/5090 (sm_120)

License

  • summer_holiday.lic — Evaluation license (16 GPUs, expires 2026-08-31)

Key Changes

  • GPU architecture flags table expanded (Kepler through Blackwell 12.0)
  • Two-architecture build matrix for libnep_gpu.so
  • Updated examples job scripts
  • CUDA ABI matching guidance

Important Notes

  • The .so CUDA version MUST match the runtime libcudart version (11.x .so needs CUDA 11.x, 12.x .so needs CUDA 12.x).

For commercial licenses, contact matpl@pwmat.com.

Assets 8
Loading