Hi, I've noticed a very strange behavior of SIESTA (trunk-364) with Intel 12 comiler a.k.a. "Composer 2011". The calculations get stuck after completion, the job doesn't terminate, and, mysteriously, even the CPU utilization on the compute nodes remains at the same level as during the actual calculations. I'm seeing this on two different machines, one with MVAPICH and one with OpenMPI, both compiled with -O3 and linked with MKL.
So, the output of the job reaches this point: ... > coxmol: Writing XMOL coordinates into file cap.xyz > > > siesta: Program's energy decomposition (eV): > ... > ... > siesta: Final energy (eV): > ... > ... > siesta: Electric dipole (a.u.) = 0.000032 0.000007 0.000000 > siesta: Electric dipole (Debye) = 0.000081 0.000018 0.000000 --- and then nothing happens until the job time runs out and the queue manager kills it. Has anybody encountered this type of behavior? Thanks in advance, Vasilii
