DEM-Engine documentation

DEM-Engine (DEME) is a GPU-accelerated discrete element method solver with a Chrono-like C++ API and Python bindings distributed as deme.

This documentation covers installation, the simulation model, the Python and C++ interfaces, and the constraints that matter when running JIT-compiled CUDA kernels.

Getting started

• Installation
  • System requirements
  • Python package
  • Build a wheel from a checkout
  • C++ build
  • Install the C++ library
  • Development and release packaging
• Quickstart
  • Python
  • C++
• Core concepts
  • Setup and runtime
  • Dynamic and kinematic workers
  • Owners and members
  • Families and prescriptions
  • Frames and quaternions
  • Runtime CUDA compilation
• Mesh particles and combined owners
  • Contact scope
  • Mesh templates
  • Rigid combined bodies
  • Mesh patches
• Examples and co-simulation
  • Run a C++ demo
  • Choose a starting point
  • Co-simulation
• Interactive visualization
  • Recommended scope
  • Writing a responsive visualization loop
  • Performance tradeoffs
  • Inspection controls
  • Scene/frame interface
  • Verification
  • Linux build requirements
  • Postprocessing with ParaView
• Troubleshooting
  • CUDA or PTX mismatch
  • Runtime JIT compilation cannot find headers
  • Import works but initialization fails
  • Conda GLIBCXX errors
  • Stale runtime kernels
  • Persistent Jitify startup cache
  • Custom force models

Interfaces

• Python interface
  • Python quickstart
  • Solver lifecycle
  • CUDA device selection
  • Retrieving simulation data
  • Python API overview
  • Python API reference: deme
  • Import and compatibility
  • Version
  • API conventions
  • Reference status
• C++ API reference
  • Solver
  • Tracking and inspection
  • Materials and force models
  • Geometry and initialization
  • Sampling

Development

• Developer architecture
  • Public API boundary
  • CUDA and JIT kernels
  • Synchronization
  • Adding Python bindings
• Internal type codes for debugging
  • Contact types
  • Geometry and owner codes
  • Analytical geometry subtypes
  • Reading IDs and debugger output
• Building the documentation
  • Prerequisites
  • Standalone build
  • Regenerating the Python API reference
  • CMake target
  • Documentation policy
• Python wheel maintenance
  • Build the complete supported Python matrix
  • Build release wheels with cibuildwheel
  • Automated wheel builds
  • Publish the deme distribution
  • Wheel portability
• Hosting the documentation
  • Recommended: GitHub Pages
  • Alternative: Read the Docs Community
  • Maintaining the source tree
• Project information
  • Community
  • Licensing and contributors
  • Citation