[gmx-users] 3.3 parallel runs - Lincs/PME/sorting errors?

[Alan Dodd]

Further to my original post, I've tried further
combinations of old/new grompp/mdrun and numbers of
nodes.  New mdrun is always less stable (produces
large amounts of LINCS errors, as opposed to none from
old mdrun) and often crashes with new grompp.  In
addition, I have one old .tpr file which runs
perfectly with old mdrun, but crashes after 3500 steps
with new mdrun.  So my question is, basically, what's
going on?  I've noticed new mdrun seems to run up to
50% faster on the same tpr as old mdrun, has this
improvement just been at the expense of stability, or
is mdrun genuinely bugged?

Thanks.

Alan Dodd (University of Bristol)


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com