The only deallocation performed by node 0 is an array containing the atomic species, I suspect you see the same error no matter the number of cores used?
It could mean a couple of things, the electrode TSHS is not fully populated (i.e. erroneous), you did a Gamma-point electrode calculation, or that there is an actual error in the communication. Can you please supply your electrode and transiesta fdf files. Could you please try and goto line 1055 of file Src/m_ts_electrode.F90 and add this (right before the "call memory" line): print*,allocated(isa),isa and recompile/run and return the output to me? Kind regards Nick 2013/10/10 zgp121 <[email protected]> > ** > ** > Dear SIESTA users: > > When I use 24 cores to run the TranSIESTA calculation using trunk 428, I > got the error like this: > > /var/spool/slurmd/job1321828/slurm_script: line 12: 19218 Bus > error mpirun transiesta < trans-0.0.fdf > trans-0.0.log > > And I also find the following message in trans-0.0.log: > > Creating Green's function file for: left > > WARNING: alloc-realloc-dealloc name mismatch > Name: elec_HS unknown > Size: -588. Bytes > Node: 0 > Subsequent mismatches will not be reported > > The system of the HPC is Ubuntu. And in the compiliation, Ifort(10.1.012), > MKL(10.0.2.018), OpenMPI(1.4.5) are used. > > Can anyone figure out what is the problem? > > Guangping Zhang > 2013-10-10 > ------------------------------ > ** >
