GROMACS upcoming 5.1.3 Release Notes

These release notes document the changes that have taken place in GROMACS since version 5.1.2 to fix known issues.

• Fixes where mdrun could return incorrect results
  • Fixed reading of bonded tables
  • Fixed bug with flat-bottomed position restraints and MPI and OpenMP
  • Fixed multi-sim + DD + simple distance restraints
  • Fixed OpenCL error with empty domains
  • Handled partially frozen and constrained atoms
  • Fixed OpenCL error with empty domains
  • Fix compile error for IBM VMX on ppc64
  • Correct nrdf for 1D/2D systems
  • Avoid assertion failure in pme_loadbal_do
  • Fix incorrect CMAP parsing
  • Changed norm(...) > 0 condition to norm2(...) > 0 in init_ekindata()
  • Fix excessive list splitting
  • Fix Ewald 3DC dipole with Verlet scheme
  • Remove restriction on dihdraltypes order
  • Fix tabulated wall potential
  • Correct mdrun tMPI (non-)parallel error message
• Fixes for mdrun performance issues
  • Added support for CUDA CC 6.0/6.1
  • Re-enable LJ-PME + Lorentz-Berthelot on GPUs
• Fixes for gmx tools
  • Fixed gmx dos index handling
  • Prevented gmx pdb2gmx writing to unallocated memory making specbonds
  • Mark optional input/output files in gmx order
  • Removed broken memory reallocation in gmx dielectric
  • Fix non-normalization of FFT autocorr
• New consistency checks
  • Added QMMM checks
• Build system enhancements
  • Ignore stderr when detecting SIMD and build info
  • Make more CMake variables advanced
  • Fix SIMD detection on new AMD AVX CPUs without fma
  • Warn about OpenMPI 1.8.6
• Improved documentation
  • Fixed constants, units and conversion factors
  • Corrected value of electric conversion factor f in the manual
  • Fix mdp label generation
  • Spelling fixes
  • Improved/enlarged Fig. 4.9 in manual for better readability
  • Re-generated EPS for figure 4.8
  • Fixed typo in Fourier dihedral equation
  • Updated reference manual text about PME tuning
  • Documented GMX_GPU_APPLICATION_CLOCKS
  • Make both ftp and http links for source download
• House-keeping that reduces users’ problems
  • Reverted to old behavior for handling SETTLE errors
  • Properly reset CUDA application clocks
  • Used sysconf(_SC_NPROCESSORS_ONLN) better on x86
  • Prevented use of mdrun -maxh -multi
  • Ignore empty GMXLIB environment variable
  • Fixed large file issue on 32-bit platforms