1 GROMACS 2018.3 release notes
2 ----------------------------
4 This version was released on August 23, 2018. These release notes document
5 the changes that have taken place in GROMACS since version 2018.2, to fix known
6 issues. It also incorporates all fixes made in version 2016.5 and
7 earlier, which you can find described in the :ref:`release-notes`.
9 Fixes where mdrun could behave incorrectly
10 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
12 Multi-domain GPU runs can no longer miss pair interactions
13 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
15 With systems with empty space in the unit cell, GPU runs with domain
16 decomposition would not compute LJ and Coulomb interactions between
17 domains when there we no interactions between domains on a rank at some
20 - This bug only affects simulations running on GPUs with domain decomposition
21 and containing empty regions of space that can lead to domains being empty.
22 - Possible observations of this bug may have been seemingly random failures
23 of calculations that where not reproducible when restarting a simulation
24 from a checkpoint file, as the domain would then again be filled properly
25 if interactions are present at the beginning.
26 - It is unlikely that this bug will have unnoticed effects on all but
27 very short simulations, as the missing interactions will inevitable lead
28 to simulation instability and crashes.
29 - If a simulation that crashed due to this bug is restarted it can contain
30 a small region around the crash that will be unphysical due to some
31 interactions not being calculated just before the crash itself.
33 **This is a critical fix and users of 2018.x series that run on GPUs should
34 update to this point release**
38 Fix Conjugate Gradient assertion failure at end of minimization
39 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
41 When the final step coincided with a coordinate output step,
42 conjugate gradient minimization would exit with an assertion failure
43 instead of writing confout.gro.
47 Multi-domain Conjugate Gradient minimimization no longer segfaults.
48 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
52 Fix pairlist buffer with Brownian Dynamics
53 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
55 With Brownian Dynamics and bd-fric > 0, the Verlet pairlist buffer would
56 be determined with incorrect masses for constrained atoms and virtual
57 sites. This would lead to a too small buffer for typical atomistic
58 systems with constraints.
62 Avoid "atom moved to far" errors
63 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
65 The introduction of the dual pair list has led to larger nstlist values,
66 which leads to larger atom displacements between domain decomposition
67 steps. This has made it more likely that the errors
68 "An atom moved too far between two domain decomposition steps" and
69 "N particles communicated to PME rank M are more than 2/3 times the cut-off
70 out of the domain decomposition cell ..." appear for stable systems.
71 Now atom displacements are correctly taken into account when determining
72 the minimum cell size, so these errors should only appear for unstable systems.
76 grompp now checks that pull groups are not close to half the box size
77 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
79 Pull groups that use a reference atom for periodic boundary treatment
80 should have all their atoms well within half the box size of this reference.
81 When this is not the case, grompp will issue a warning.
85 Fixed segmentation fault in mdrun with QM/MM ONIOM scheme
86 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
90 Correctly specified that PME on GPUs is only supported for dynamical integrators
91 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
93 Previously PME on GPU support could run (but fail) for energy
94 minimization and normal-mode analysis runs.
98 Fixes for ``gmx`` tools
99 ^^^^^^^^^^^^^^^^^^^^^^^
101 Fixed syntax error in make_gromos_rtp.py
102 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
106 Fix gmx solvate topology updating
107 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
109 Removed hard coded solvent names to allow updates to topology based on
110 solvent molecule information. Also allows updating with multiple solvent
115 Fix bfactor output error caused by fix for :issue:`2511`
116 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
117 The fix for the PQR file output broke the output of bfactors from other tools.
121 Made sure that gmx rms can skip values
122 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
123 When requested to skip values, gmx rms would still output all values despite
124 the option. Now it only outputs values that are requested to be processed.
128 Fix trjconv when not providing structure file
129 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
131 trjconv would fail with a segmentation violation when running without any structure
132 file due to incomplete initialization of the topology data structure. This fix adds
133 the missing checks that prevents the failure.
137 Fix enforced rotation energy output
138 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
141 Fixes to improve portability
142 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
144 Fix nvcc host compiler check triggering
145 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
150 Report up to date hwloc version information
151 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
156 Disable single compilation unit with CUDA 9.0
157 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
165 Avoid aborting mdrun when GPU sanity check detects errors
166 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
171 Improve OpenCL kernel performance on AMD Vega GPUs
172 """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
173 The OpenCL kernel optimization flags did not explicitly turn off denorm handling
174 which could lead to performance loss. The optimization is now explicitly turned
175 on both for consistency with CUDA and performance reasons.
176 On AMD Vega GPUs (with ROCm) kernel performance improves by up to 30%.