New features

• A subset of the system (rather than the whole system) can be coupled to the temperature bath both with Berendsen and Nose-Hoover coupling. It is done by setting the tau_T to zero for the groups for which you do not want coupling in the mdp file for these groups. A subsection has been added in the printed manual.
  
• Successive over relaxation for shake (in order to reduce the number of iterations for shake) is now an mdp option (Shake-SOR), turned off by default.
  
• Merged mdrun with nmrun and xmdrun. A new integrator (nm for normal mode analysis) was introduced in the mdp file. This reduces code duplications.
  
• Code cleanups in the shake stuff. Obsolete old code was removed, and complexity reduced.
  
• Added a new analysis tool g_clustsize that computes the size of clusters, based on distance.
  
• Support for assembly loops on MacOS X (using altivec)
  
• Support for double precision SSE2 assembly loops on the Pentium IV
  
• Support for the Intel compilers, both on x86 and ia-64 platforms
  
• Inner loops that do not compute forces have been written. For really fast Monte Carlo, but there is no integrator yet.
  
• Cleanup of the EM initialization code
  
• Rotation around the center of mass can be removed, also in groups and on parallel computers
  
• Support for orientational restraints
  
• Distance restraint code has been improved. Ensemble averaging is now possible over non-interacting subsystems. 
  
• tpx files now contain the atom types. This could be very useful for users programming their own analysis tools, and is already used in e.g. g_sas (solvent accessible surface). 
  
• g_sas now takes peridicity into account and makes an estimate of the free energy of solvation based on the solvent accessible area.
  
• Waters (settles) are made whole now (i.e. molecules are kept together), this is especially important for trjconv.
  
• Charge and dipole corrections for Ewald long-range electrostatics
  
• Pseudo-2d Ewald long range electrostatics by using correction terms (a.k.a. EW3DC). This is selected with the new ewald_geometry option in the mdp file.
   

Bug fixes

• Bug was fixed in conjugate gradients minimization with PME. In 3.0.x the long range Coulomb was not taken into account at all.
  
• A bug was fixed in trjconv which might explain problems people have had when fitting.