[gmx-developers] Nonconsecutive shake groups?

[Shirts, Michael (mrs5pt)]

Just checking - currently, is there no way to have constraints on blocks of
nonconsecutive atoms?  For example, with the following topology, with
constraints set to h-bonds:

[ atoms ]
;   nr  type  resi  res  atom  cgnr     charge      mass       typeB
chargeB
     1   ca     1   MET    C1    1     -0.03810     12.01000
     2   ca     1   MET    C2    2     -0.02890     12.01000
     3   ca     1   MET    C3    3     -0.17620     12.01000
     4   ca     1   MET    C4    4     -0.23470     12.01000
     5   ca     1   MET    C5    5     -0.17410     12.01000
     6   ca     1   MET    C6    6      0.15910     12.01000
     7    c     1   MET    C7    7      0.64940     12.01000
     8   c3     1   MET    C8    8      0.12850     12.01000
     9    o     1   MET    O1    9     -0.54700     16.00000
    10   oh     1   MET    O2   10     -0.49020     16.00000
    11   os     1   MET    O3   11     -0.44230     16.00000
    12   ha     1   MET    H1   12      0.16160      1.00800
    13   ha     1   MET    H2   13      0.15640      1.00800
    14   ha     1   MET    H3   14      0.15790      1.00800
    15   ha     1   MET    H4   15      0.14070      1.00800
    16   h1     1   MET    H5   16      0.06060      1.00800
    17   h1     1   MET    H6   17      0.04650      1.00800
    18   h1     1   MET    H7   18      0.04620      1.00800
    19   ho     1   MET    H8   19      0.42460      1.00800

it started to generate a lot of "Double sids" errors as soon as mdrun
started, apparently because the first constraint block was interpreted as
being from 1 to 12.

1) it would be nice to be able to have constraint blocks including
nonconsecutive atoms (I think I had a conversation about this with David van
der Spoel at some point, but can't find it in the gmx-developer's archive),
since some molecule manipulation programs output chemicals in this format,
and it might be a pain to rewrite it?

2) at the least, grommp should give a more informative warning, instead of
getting to mdrun with lots of "double sids" error that don't describe what's
going on.

3) or am I doing something wrong?
  

Thanks!
~~~~~~~~~~~~
Michael Shirts
Assistant Professor
Department of Chemical Engineering
University of Virginia
michael.shirts at virginia.edu
(434)-243-1821