Features

• Newtonian and Chebyshev polynomial propagators of arbitrary order
• Verlet integrator implemented as reference
• Different expression evaluation strategies implemented as different classes
• Use of different backends, including mpmath, numpy, theano and tensorflow
• Arbitrary precision arithmetics using the mpmath backend
• Support arbitrary pair-interaction forcefields (Coulomb, Lennard-Jones, etc.) for arbitrary number of particles
• Periodic boundary conditions at distance matrix level and at pair potential level
• Utilities for static expression analysis, and for performance and accuracy measurements

Known issues

• Sometimes long startup / compilation times

TODOs

• Comprehensive documentation
• Dissipative dynamics and thermostats
• Potential truncation (potential cutoffs)
• Reduce number of operations using symmetries and expression manipulations
• Reuse constant and conform expressions to save memory and accelerate preparation phase
• Parallelization on heterogeneous clusters with GPUs