From: Jim Phillips (jim_at_ks.uiuc.edu)
Date: Mon Jul 18 2011 - 15:35:28 CDT
This is basically a "Bad global bond count" error, which would normally
only be a warning during minimization except it is being caught by some
less forgiving code because while (in this case) patch 1005 has a proxy on
the node it is not one of the patches belonging to the bonded compute.
This situation is more likely when there are fewer atoms per core, which
is why you see it on a Blue Gene.
I've checked in a fix that will print a warning and skip the tuple but
let the more forgiving checksum code catch the error later.
-Jim
On Sat, 18 Jun 2011, JC Gumbart wrote:
> I'm getting this error when running on a Blue Gene, but not on a normal Linux cluster. I also tried the memory optimized version but still got the same error. The system size is about 350k atoms.
>
> TCL: Minimizing for 2000 steps
> Warning: Tuple with atoms 11563(1003) 43551(1005) missing patch 1005
> FATAL ERROR: Patch needed for tuple is missing.
>
> Warning: Tuple with atoms 32385(958) 37155(960) missing patch 960
> FATAL ERROR: Patch needed for tuple is missing.
>
> I've seen it come up before in the list, but with no resolution. Any ideas?
>
> Thanks,
> JC
>
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:20:35 CST