GROMACS 2018.2 release notes¶
This version was released on June 14, 2018. These release notes document the changes that have taken place in GROMACS since version 2018.1, to fix known issues. It also incorporates all fixes made in version 2016.5 and earlier, which you can find described in the Release notes.
Fixes where mdrun could behave incorrectly¶
Prevented OpenCL timing memory leak¶
When using OpenCL builds and timing, a memory leak would lead to all system memory being used up.
Fixed MPI error after constraint failure during energy minimization¶
Fixed moving frozen atoms with constraints¶
Frozen atoms which also had bond constraints could move.
Fixed COM removal moving frozen atoms¶
When frozen atoms were part of center of mass motion removal groups, they could accumulate momentum and move.
Fixed AWH too infrequent checks for covering¶
For multidimensional AWH grids with many points relative to the number of samples required for covering the grid, the detection of covering could be delayed because of too infrequent checks.
Fixed AWH continuation consistency checks¶
Some kinds of changes upon restarts are now disallowed, as intended.
Fixed AWH awh-nsamples-update value checking¶
Zero is now forbidden, as intended.
Fixes for gmx tools¶
Fixed buffer overflow in grompp warnings¶
When grompp issued a warning or error with a filename/path of more than 255 characters, a buffer overflow would occur. This could also happens during make check/test.
Fixed infinite loop in gmx solvate¶
When provided with a PDB file that had no box information for the solvent, gmx solvate could be stuck in an infinite loop. Fixed by disallowing empty boxes for solvent PDB files.
Fixed enemat when the .edr file had no matching energy groups¶
Fixed PQR file output¶
PQR files from gmx editconf violated the standard for the format because they were always written in fixed format. This commit fixes the issue by introducing a different output method for PQR files that follows the standard.
Fixed crash in gmx solvate¶
gmx solvate would crash due to memory corruption when using multiple solvent molecule types.
Added check for unallowed periodic setups¶
Long distances between atoms in bonded interactions could lead to incorrect periodicity removal. In such cases an inconsistent shift message was printed, but the run or analysis was not terminated and other, confusing, errors could be issued. Now an informative fatal error is issued.
Fixes to improve portability¶
Fixed CUDA compilation on Windows.¶
Fixed SIMD support for POWER systems in double precision with gcc 8¶
Fixed possible illegal instruction on KNL with Intel compiler¶
Miscellaneous¶
Information message about OMP_NUM_THREADS now sent to log file¶
Made it easier to track this information by writing it to the log file in a clear way, rather than to stderr.
Fixed inadvertent disabling of SIMD version of the integrator¶
Fixed a bug so the SIMD version of the leap-frog integrator is chosen, when possible. This may improve performance.
Fixed own FFTW builds on certain AVX2/AVX512 hardware¶
Version 3.3.8 of FFTW fixes some known gcc-8 errors for AVX2 by removing the fast-math flag, and it also appears to fix an issue with failed unit tests on AVX512-capable hardware, so we have bumped the version we download to 3.3.8.
Switched to using more standard CMake variables for installing on GNU systems¶
GnuInstallDirs.cmake is a better approach.
Several documentation and output improvements¶
Updated top-level README file for latest GROMACS core publication.
Reporting about GPU detection has improved.
gmx mindist -pidocs improved.Docs for mdp options relating to bonds improved.
Fixed various typos.
Removed a leftover mention of the twin-range scheme.
gmx trjconv -ndecdocs improved.