GROMACS 2020.5 release notes¶
This version was released on January 6th, 2021. These release notes document the changes that have taken place in GROMACS since the previous 2020.4 version, to fix known issues. It also incorporates all fixes made in version 2019.6 and earlier, which you can find described in the Release notes.
Fixes where mdrun could behave incorrectly¶
Fix mdrun writing zero dH/dlambda and foreign lambda energies before checkpointing¶
With free-energy runs with separate-dhdl-file=no and nstdhdl not a multiple of nstenergy, mdrun would write zeros for dH/dlambda and foreign energies to the energy file for steps between the last energy frame and the checkpoint. This would lead to errors in free-energy estimates which could go unnoticed as values only deviate for a few steps.
Fixed bugs with COM pulling and domain decompostion with weight or >32 ranks¶
When using COM pulling and domain decomposition, the results would be incorrect when using relative weights per atom or when using more than 32 DD MPI ranks. This would usually lead to crashes or obviously wrong results.
Fixed conserved energy for MTTK¶
When using pcoupl=MTTK and tcoupl=nose-hoover, the calculated conserved
energy was incorrect due to two errors dating back to GROMACS 4.6 and 2018,
respectively. As a result, all reported conserved energies using this
combination of temperature and pressure coupling algorithms in any GROMACS
version since GROMACS 4.6 are likely to be wrong. Note that these errors did
not impact the dynamics, as the conserved energy is only reported, but never
used in calculations. Also note that this bug only affects this exact
combination of temperature / pressure coupling algorithms.
Fixed conserved energy for Nose-Hoover¶
When using tcoupl=nose-hoover and one or more temperature groups with
non-integer number of degrees of freedom, the calculated conserved
energy was incorrect due to an error dating back to GROMACS 2018.
Reported conserved energies using Nose-Hoover temperature coupling and
non-integer number of degrees of freedom since GROMACS 2018 are likely to
be slightly off. Note that this error does not impact the dynamics, as the
conserved energy is only reported, but never used in calculations. Also note
that this will only be noticeable when using small temperature groups or
small systems.
Fixed kinetic energy and temperature reporting for MTTK¶
When using pcoupl=MTTK and tcoupl=nose-hoover, the reported kinetic
energy and temperature were very slightly off. The integration of the
temperature coupling trailed the reporting by half a time step. Note that
these errors did not impact the dynamics, as the quantities were correctly
integrated and only wrongly reported. Also note that the difference is so
small that it is unlikely to have been significant for any application
except for rigorous algorithm validation. Finally, note that this bug
only affects this exact combination of temperature / pressure coupling
algorithms.
Fix pull error message with angles and dihedrals¶
The COM pull code could print incorrect pull group indices when mdrun exited with an error about a too long pull distance in angle and dihedral geometries.
Fix numerical issues in expanded ensemble¶
When performing simulated tempering or expanded ensemble simulations with changes in the Hamiltonian that were too large, then Monte Carlo proposals to states that were sufficiently unlikely would underflow, causing division by zero errors. This was fixed by numerically hardening the logical flow so that such proposals would be rejected instead.
Fix incorrect electric field strength with applied electric field¶
The electric field generated by the electric field module would be incorrect when used together with domain decomposition due to an error with indexing the field to all atoms instead of just those on the current domain.
In overlap regions between domains, which have the thickness of the pairlist cut-off distance, the electric field would be doubled (or more with 2D or 3D domain decomposition).
To validate if a simulation has been affected by the issue, users should calculate the actual potential across the simulation box using the Poisson equation. If this potential agrees with the one provided as the input, a simulation was not affected.
Fixes for gmx tools¶
Improve CHARMM support in gmx do_dssp¶
Fix non-funtioning gmx h2order -d option¶
The gmx h2order tool would always take the normal along the z-axis.
Fix pull group index handling¶
The pull code would not validate its index groups correctly, leading to infinite loops or assertions being triggered at grompp time.