GROMACS 2018.1 release notes ---------------------------- This version was released on March 21, 2018. These release notes document the changes that have taken place in GROMACS since the initial version 2018, to fix known issues. It also incorporates all fixes made in version 2016.5 and earlier, which you can find described in the :ref:`release-notes`. Fixes where mdrun could behave incorrectly ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Fixed leap-frog integrator with Nose-Hoover T coupling and Parrinello-Rahman P coupling """"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" With Parrinello-Rahman P coupling active, when applying Nose-Hoover T coupling at an MD step where no P coupling occured, the update phase could use outdated or garbage coupling data. Such simulations with :mdp:`nsttcouple` equal to :mdp:`nstpcouple` are unaffected by this issue, so few users will be impacted by this. Simulations using other coupling algorithms are unaffected. :issue:`2418` Used SIMD bondeds without perturbed interactions """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" In free-energy calculations that lacked bonded interactions between perturbed atom types, the SIMD-accelerated bonded functions were inadvertently disabled. This has been enabled, which will improve the performance of some kinds of free-energy calculations. Fixed bonds whose displacement was zero """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" We should allow overlapping atoms in harmonic bonds. But the former code would cause a floating point exception and incorrect free-energy derivatives. Fixed centre-of-mass motion removal on part of the system """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" COMM removal requested for part of the system acted on the whole system. :issue:`2381` Fixed multi-simulations with multiple ranks per simulation """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" These used to crash or hang mysteriously before the simulation would start. :issue:`2403` Improved inter-simulation signalling implementation """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Reduced communication overhead with either many simulations or many ranks per simulation. Fixed FEP calculations with SHAKE """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" All SHAKE + FEP calculations accumulated wrong values to dH/dl output, but in some cases the result will look the same. :issue:`2434` Fixed handling of mdp ``define`` statement assigning preprocessor values """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Now .mdp files can configure the topology with values, as originally intended, e.g. ``"define = -DBOOL -DVAR=VALUE"``. :issue:`2392` Prevented log file energy average printing dividing by zero """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" If very few simulation frames have computed energies, then there may be insufficient data for averages. If so, skip the average printing entirely. :issue:`2394` Correctly set cutoff modifiers in forcerec """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" The cutoff modifiers were not copied from interaction_const_t to forcerec_t which meant only the generic kernels were used with the group scheme. This fix will restore the performance of the group scheme. :issue:`2399` Fixed box scaling in PME mixed mode using both GPU and CPU """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" :issue:`2385` Re-enabled GPU support with walls and 1 energy group """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" With a single non-bonded energy group and walls, we can now use a GPU for non-bonded calculations. Removed tumbling ice-cube warning with SD integrator """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" With SD, there is friction, so ice cubes will not tumble. Fixed assertion failure in test-particle insertion """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Erroneous logic in the TPI meant that it always failed without producing any result. :issue:`2398` Avoided mdrun echoing "No option -multi" """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" mdrun would print as many messages "No option -multi" as there are MPI ranks to stderr. Also updated -multi to -multidir in an error message. :issue:`2377` Improved mdrun handling when GPUs are present but unavailable """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" :issue:`2415` Fixed crash with AWH and awh1-equilibrate-histogram=yes """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" When running AWH with awh1-equilibrate-histogram=yes and multiple MPI ranks, the simulation would segmentation fault. :issue:`2436` Fixed issues with AWH and bias sharing """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" When sharing AWH biases between multiple simulations, there were four issues. An MPI error would occur when an individual simulation would use more than one rank. The deconvoluted PMF would be garbage (but the sampling was correct). with more than 32 MPI ranks for an individual simulation, an error about a coordinate being 0 could occur. And continuation from checkpoints could be refused. :issue:`2433` :issue:`2439` :issue:`2441` :issue:`2444` Fixed virial with AWH and domain decomposiion """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" When running AWH with domain decomposition, the AWH/pull virial contribution would be multiplied with the number of MPI ranks. Fixed restart bug with pull geometry direction-periodic """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" With COM pulling with geometry direction-periodic, (only) at the step of continuing from checkpoint the closest PBC image would be used instead of the of the one closest to the reference value. This could lead to a sharp spike in the pull force at the continuation step. :issue:`2446` Fixes for ``gmx`` tools ^^^^^^^^^^^^^^^^^^^^^^^ Added check in grompp to avoid assertion failure """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" With an mdp file with a parameter present with both the current name and the old name which automatically gets replaced, an assertion would fail. Now a fatal error is issued. :issue:`2386` Fixed grompp net charge check """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Use of multiple non-consecutive blocks of a moleculetype now works correctly. :issue:`2407` Fixed issue with adding selection groups for TNG output """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" When there were more molecule blocks than molecule types in the topology, the output was wrong. Fixed help text and functionality of ``pdb2gmx -missing`` """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" This now permits dangling bonds at termini, which is occasionally useful. Fixes to improve portability ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ PME on Fermi-era GPUs on large systems now works """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" On older GPUs, it was possible to run into a hardware size limitation that has now been fixed. :issue:`2409` GoogleTest death tests are now used in a more portable way """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Tests for GPU utility functionality are now more robust """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Non-GPU builds, and GPU builds that find incompatible or otherwise unavailable devices will pass the tests in the manner intended. :issue:`2405` Used more portable python shebangs """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Per https://www.python.org/dev/peps/pep-0394/#recommendation, we should use env, and point it at python2. When we either make them 2/3 or just-3 compatible, this should change. Some distros (notably Arch Linux) already point python at python3 so we should choose to be explicit, and thus somewhat portable. :issue:`2401` Added work-around for GCC 5.3 targetting AVX512 hardware """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" GCC 5.3 has bug in overload resolution causing the AVX512 and scalar function to become ambiguous. Used isfinite unambiguously """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Patch provdied by Veselin Kolev to quiet some compiler warnings. :issue:`2400` Worked around gcc-6 bug in tabulated group non-bonded kernels """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" With the gcc-6 compiler, AVX and -O3, which is the default, the tabulated non-bonded kernels of the (deprecated) group cutoff-scheme produced incorrect energies and forces. The errors are so large that they could not have caused latent issues. :issue:`2424` Detected correct AMD Zen SMT topology """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" On recent AMD Zen processors, hardware thread detection and pinning handling have been fixed, improving performance. :issue:`2388` Fixed POWER VSX SIMD usage for upcoming gcc version 8 """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" :issue:`2421` Fixed clang 6 with CUDA 9 """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Permits builds for sm_70 and may work around an issue with sm_37 :issue:`2443` Miscellaneous ^^^^^^^^^^^^^ Made multi-atom TPI reproducible with different compilers """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" Documentation enhancements """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" In particular, for handling options to mdrun relating to GPUs and running mdrun with good performance.