Released in NAMD 2.10

The following features have been released in NAMD 2.10. Any bug fixes will appear in the nightly build version on the download site. Any documentation updates will appear in the Nightly Build User's Guide (online or 1.7M PDF) and release notes.

Network topology adaptation for CRAY XE/XK and IBM Blue Gene/Q

This advance allows the largest simulations to scale to the full Blue Waters and Titan machines, as described in our SC14 paper.

CRAY XC30 and IBM Blue Gene/Q native network layers

The new Charm++ machine layers are gni-crayxc, pamilrts-bluegeneq, and pami-bluegeneq.

Multiple-copy algorithm support for native network layers

Replica exchange in NAMD 2.9 required the underlying Charm++ library to be based on a modified MPI machine layer. NAMD 2.10 utilizes the new Charm++ 6.5.0 partitioning facility, which is available on all machine layers based on the Converse "LRTS" low-level runtime system. LRTS machine layers include gemini_gni-crayxe, gni-crayxc, pamilrts-bluegeneq, verbs, netlrts, and mpi. Existing NAMD 2.9 replica-exchange Tcl scripts should work unmodified in NAMD 2.10. This development was funded by NCSA/Illinois Enhanced Intellectual Services for Petascale Performance(NEIS-P2).

Lambda-exchange multiple-copy alchemical free energy calculations

Described in Generalized scalable multiple copy algorithms for molecular dynamics simulations in NAMD. Computational Physics Communications, 185:908-916, 2014.

String method with swarms-of-trajectories transition pathways

Also described in Generalized scalable multiple copy algorithms for molecular dynamics simulations in NAMD. Computational Physics Communications, 185:908-916, 2014.

Atom coordinate/velocity exchange for multiple-copy algorithms

Useful when lambda-exchange is not sufficient or for duplicating walkers as described in Multiple-replica strategies for free-energy calculations in NAMD: Multiple-walker adaptive biasing force and walker selection rules. Journal of Chemical Theory and Computation, 2014.

Tcl scripting command to reinitialize atoms from output files

Used in, e.g., adaptive multilevel splitting.

Tcl scripting configuration file parameter introspection

Tcl scripting trajectory and restart output file switching

Multilevel summation method for long-range electrostatics

Described here and in Multilevel summation method for electrostatic force evaluation. Journal of Chemical Theory and Computation, 2014. In press. (journal)

Normalized external electric field to avoid pressure artifacts

An external electric field that causes ions to diffuse repeatedly across the periodic cell leads to infinitely decreasing reported potential energy and infinitely increasing reported pressure. Enabling eFieldNormalized indicates that the eField vector has been scaled by the cell basis vectors, thus indicating the voltage drop across the cell in units of kcal/(mol e). The eField vector is then scaled by the reciprocal lattice vectors at each timestep. As a result the eField energy is independent of the cell dimension and hence eField forces do not contribute to the pressure calculation, thus eliminating the pressure increase, but not the potential energy decrease.

Improved scalability for collective variables, Tcl forces, etc.

Collective variables module improvements

Drude polarizable force field improvements

SMOG structure-based potential support

Minimizer stability and performance improvements

The conjugate gradient minimization algorithm has been enhanced to (as far as we know) eliminate the tendency to induce non-physical conformations in multi-million atom systems, or for smaller systems with CUDA builds.

Various improvements for 100-million-atom simulations

Update to CUDA 6.0 with Kepler GPU optimizations

GPU-accelerated particle-mesh Ewald calculation

Only enabled by default for PMEInterpOrder 6 or 8, which are more expensive than the typical order 4.

Intel Xeon Phi coprocessor support

64-bit Windows shared-memory single-node port with CUDA support

32-bit and 64-bit ARM processor Linux port with CUDA support

CUDA support for POWER processor Linux port

Enhanced performance and scalability