Home

AMR — Parallel Adaptive Mesh Refinement

A linear (pointerless) Morton/Z-order quadtree (2D) / octree (3D) AMR engine. Three C++ backends — OpenMP, MPI, CUDA — share one Morton core and one algorithm. Doubles as the mesh backend for the adaptive DIC/DVC solver.

• 📦 Repository: https://github.com/nikbott/amr (trunk: dev)
• 📖 Build & layout: README
• 📚 Citations: REFERENCES.md

Design at a glance

• Sorted-leaf (code, level) array; refine / 2:1-balance / coarsen operate on it.
• p4est/t8code family ([BurstWG2011], [Holke2018]); non-conforming constraints [CDK2019].
• Headline: active-front 2:1 balance (CUDA + OpenMP), byte-identical, ~1.84× sphere.

Pages

• Getting started — build + run a demo.
• Backend Validation — SOTA validation of the omp / mpi / cuda backends.
• GPU Balance Notes — GPU balance algorithm notes.

Project

• Issues · Milestones
• CONTRIBUTING · CODE_OF_CONDUCT · SECURITY

Wiki pages mirror docs/ for browsing; the in-repo copies are canonical.