peclet 0.6.0

Family release: peclet-flow 0.4.0 (BREAKING: pnm split out), NEW peclet-pnm 0.1.0 (pore-network extraction + DNS network flow), peclet-dem 0.4.0, peclet-voro 0.4.0, peclet-morton 0.2.1; extras: peclet-core 0.5.0 (mpi), peclet-coupling 0.3.0 (cfd-dem).

Peclet v0.2.0

Feature release: multi-rank Python API for flow/voro + HPC MPI containers (Snellius CUDA multi-GPU, TU/e SMM multi-node CPU), per-site bind wrappers + submit scripts, and a weak-scaling communication benchmark. See the CHANGELOG.

Install: pip install peclet (CPU family). Docs: https://computational-chemical-engineering.github.io/peclet/

peclet v0.1.0 — first public release

First public release of peclet — a suite of GPU-accelerated, MPI-parallel codes for transport phenomena (CFD, DEM, Voronoi), on a shared Kokkos + block-decomposition foundation. MIT licensed.

Install (Python)

pip install peclet          # CPU (OpenMP) family: peclet-flow + peclet-dem + peclet-voro + peclet-morton

All six packages are on PyPI: peclet-core, peclet-flow, peclet-dem, peclet-voro, peclet-morton, and the peclet metapackage.

Highlights

• peclet.flow — incompressible cut-cell IBM Navier–Stokes on a staggered MAC grid, with a distributed (MPI) multi-rank solver exposed to Python.
• peclet.dem — XPBD discrete-element packing + distributed step.
• peclet.voro — dynamic Voronoi tessellation + distributed VoronoiHalo.
• peclet.core — the shared MPI halo + AMR octree; peclet.morton — Morton/Z-order arithmetic.
• Kokkos backends (OpenMP / CUDA / HIP); the CPU build ships as self-contained wheels.

HPC containers (GHCR)

Apptainer images for Snellius (peclet-cuda:0.1.0-sm80/-sm90) and multi-node CPU / TU/e SMM (peclet-cpu:0.1.0), with per-site MPI bind wrappers + SLURM submit scripts and a weak-scaling benchmark (benchmarks/). The LUMI/HIP image is WIP (a hipcc/lld link issue). See the Containers docs.

Docs

https://computational-chemical-engineering.github.io/peclet/