Hallo everybody,
I am Francesco Zappa, a PhD student at the Jena university. I am using a Cactus-based code and I am running simulations on Hawk cluster. Unfortunately, recent policy of Hawk have severe restrictions in terms of number of files that can be produced at the same time. The problem is that running several mpi processes generate many hdf5 output files and I would like to have them packed somehow. I have tried to use the option
CarpetIOHDF5::one_file_per_proc = "yes"
which works fine for 2D data files but it does not seem to work on 3D data files (which have the form <variable>.file_<process>.h5).
I have been aware of the fact that it exists a patch for the Carpet Thorn to have these files packed together somehow. Could you please help me with this issue?
Best regards,
Francesco Zappa
-------------------------------------
Friedrich-Schiller-Universität Jena Theoretisch-Physikalisches Institut Fröbelstieg 1, Office 219, Phone
Number 0049-3641-9-47133 D-07743 Jena