[gmx-developers] GB + Charmm problems

[David van der Spoel]

Hi,

I'm trying to run some simulations of a protein in GB solvent using the 
charmm force field. One of the simulations crashes after 2 ns with what 
seems to be a table lookup error (possibly due to strange coords):
imb F 22% step 1023500, will finish Tue Jul 13 05:33:06 2010
imb F 22% step 1023600, will finish Tue Jul 13 05:33:08 2010
imb F 22% step 1023700, will finish Tue Jul 13 05:33:11 2010
[a04c01n15:21227] *** Process received signal ***
[a04c01n15:21227] Signal: Segmentation fault (11)
[a04c01n15:21227] Signal code: Address not mapped (1)
[a04c01n15:21227] Failing at address: 0x2a92aee1a014
[a04c01n15:21234] *** Process received signal ***
[a04c01n15:21234] Signal: Segmentation fault (11)
[a04c01n15:21234] Signal code: Address not mapped (1)
[a04c01n15:21234] Failing at address: 0x2a92aee18014
[a04c01n15:21227] [ 0] /lib64/libpthread.so.0 [0x33c440e4c0]
[a04c01n15:21227] [ 1] mdrun(nb_kernel330+0x2c3) [0x6091d3]
[a04c01n15:21227] [ 2] mdrun(do_listed_vdw_q+0xa84) [0x598204]
[a04c01n15:21227] [ 3] mdrun(calc_bonds+0x261) [0x513e01]
[a04c01n15:21227] [ 4] mdrun(do_force_lowlevel+0x1045) [0x472175]
[a04c01n15:21227] [ 5] mdrun(do_force+0xddc) [0x4cb81c]
[a04c01n15:21227] [ 6] mdrun(do_md+0x617a) [0x43363a]
[a04c01n15:21227] [ 7] mdrun(mdrunner+0x11b8) [0x42d248]
[a04c01n15:21227] [ 8] mdrun(main+0x9a0) [0x435c70]
[a04c01n15:21227] [ 9] /lib64/libc.so.6(__libc_start_main+0xf4) 
[0x33c381d974]
[a04c01n15:21227] [10] mdrun [0x41e589]
[a04c01n15:21227] *** End of error message ***

Similar crashes have happened in the CMAP code where there is no check 
for index out of bounds, before using the lookup table. At one stage I 
also had a large NEGATIVE LJ energy (-1e7), maybe due to atoms coming 
vey close to each other?

Relevant MDP lines:
integrator               = md
dt                       = 0.002
implicit_solvent         = GBSA
gb_algorithm             = Still
nstgbradii               = 1
rgbradii                 = 1.5
gb_saltconc              = 0.2
rlist                    = 1.5
nstlist                  = 5
ns-type                  = grid
pbc                      = no
tcoupl                   = v-rescale
tc-grps                  = system
tau-t                    = 0.01
ref-t                    = 320

everything else is default.

Any clues how to debug this? This crash is after a few hours on 8 cores.

-- 
David van der Spoel, Ph.D., Professor of Biology
Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205.
spoel at xray.bmc.uu.se    http://folding.bmc.uu.se