Niladri, I would bet that if you used 600 Ry on a single processor the error would not exist :o). In case you are running parallel siesta serially and get no error, it becomes rather clear that the error happens from using netcdf with a parallel run - nothing whatsoever to do with mesh cutoff. There are only two solutions: either you recompile netcdf to have it give you no error when running the code in parallel, or you simply don't use it in parallel. Right now I don't remember if there's an fdf flag to disable netcdf output, but if there are none, I suggest you remove every reference to it in the arch.make file and recompile siesta without netcdf support.
Have a nice weekend, Marcos On Fri, Dec 11, 2009 at 6:51 PM, niladri patra <[email protected]> wrote: > Hi, > > I have the following error: > > > > InitMesh: MESH = 360 x 360 x 360 = 46656000 > InitMesh: Mesh cutoff (required, used) = 300.000 353.960 Ry > netCDF error: NetCDF: Invalid argument > ERROR STOP from Node: 0 > Application 1837453 exit codes: 1 > Application 1837453 exit signals: Killed > Application 1837453 resources: utime 0, stime 0 > > > > I do not understand what to do with this. I increase the meshcutoff to 600 > Ry , still the problem exist. > > But the interesting thing when I used single possessors meshcutoff > 100Ry(default) was fine for the system. But now I am using 8 possessors > (Running on 8 nodes in parallel) and I have the error. > > Is there any relation between the meshcutoff and possessors? Any suggestion > would be appreciated? > > Thanks, > > Niladri Patra >
