[Users] Problems with output in CarpetX

Erik Schnetter schnetter at gmail.com
Mon Jun 30 09:45:03 CDT 2025


Javiera

I recommend using 3D output in the way others are. CarpetX supports many parameters, but not all parameter combinations are well tested, or are intended for production runs.

(1, 2) Yes, there are two ways of 3D ASCII (tsv) output. This is very slow and is thus only intended for debugging. Therefore it doesn't have all the features needed for comfortable output in production runs.

(3) Either OpenPMD or Silo output are the preferred ways for production runs. These formats are also supported by post-processing and visualization tools. These are not file formats, but they only describe how the metadata are arranged (i.e. what names the attributes have) in the output files. The actual file formats are ADIOS2 (preferred, faster) or HDF5. OpenPMD uses ADIOS2 by default, Silo uses HDF5 by default.

I do not know why the OpenPMD/HDF5 output is not working. The error message doesn't show enough detail.

If you want HDF5 output then I recommend using Silo. This will produce an HDF5 file. I'm not sure that's what you want because, whatever you are going to do with the file, you will need to understand the metadata in the file, and thus that tool or script will need to understand either the OpenMPD or Silo layout. Otherwise your HDF5 will look like a bag of 3D arrays, each array describing a Fortran array, and it will be very difficult to find out which component of which variable and which time step on which refinement level is stored in what array: That's what the metadata describes.

On the other hand, if you want to have an easy way to read the file, then you could use e.g. the Julia or Python of C++ bindings of the OpenMPD library, and you then wouldn't need to care whether the low-level file format is ADIOS2 or HDF5.

What do you want to do with the 3D output?

-erik

> On Jun 27, 2025, at 12:54, Javiera Hernández Morales <javihernandezm at posteo.net> wrote:
> 
> Hi all,
> 
> I am starting to learn the use of CarpetX and for this I started using the WaveToyX thorn. I wanted to ask for guidance for three problems that I have encountered so far, all of them related to the output data:
> 
> 1) I wanted to get 3D data, for example in .tsv format. Here I realised that there is an ambiguous definition in the documentation for the parameter “out_tsv”, which in section 12.1 it says to control TSV output and have default value “yes”, but in section 14 it says to control the 3D output in TSV and have default value “no”. This last behaviour is the one that I encounter when running the parfiles. 
> 
> 2) When I set up “out_tsv" to “yes” and “out_tsv_every” to a number, for example = 16, I get 1D output every 16 iterations, but 3D output in every iteration, therefore I suspect that “out_tsv_every” is only controlling the frequency of the 1D output, and I cannot find a way to control the frequency of 3D output.
> 
> 3) I attempted to output data in HDF5 format but I did not succeed. I first set the parameter “openpmd_format” to “HDF5” and also set “out_openpmd_vars” to the variables that I wanted, but I got an error in the first iteration:
> 
> INFO (CarpetX): OutputGH: iteration 0, time 0.000000, run time 1 s
> INFO (CarpetX): OutputOpenPMD...
> INFO (CarpetX): Creating openPMD object...
> INFO (CarpetX):   options: 
>   {
>     "adios2": {
>       "dataset": {
>         "operators": [
>         ]
>       }
>     }
>   }
> 
> [et-juphub:699479] *** An error occurred in MPI_Comm_create_keyval
> [et-juphub:699479] *** reported by process [211877889,2]
> [et-juphub:699479] *** on communicator MPI_COMM_WORLD
> [et-juphub:699479] *** MPI_ERR_ARG: invalid argument of some other kind
> [et-juphub:699479] *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
> [et-juphub:699479] ***    and potentially your MPI job)
> [et-juphub:699473] PMIX ERROR: UNREACHABLE in file ../../../src/server/pmix_server.c at line 2193
> [et-juphub:699473] 3 more processes have sent help message help-mpi-errors.txt / mpi_errors_are_fatal
> [et-juphub:699473] Set MCA parameter "orte_base_help_aggregate" to 0 to see all help / error messages
> Simfactory Done at date: Thu 26 Jun 2025 09:19:08 AM UTC
> And I did not know how to move on from there.
> I am working on the ETK Tutorial server, in case that is important.
> Thanks in advance to anyone who took the time to read until here, I will be looking forward to any piece of advice :)
> Best,
> Javiera H.M. 
> _______________________________________________
> Users mailing list
> Users at einsteintoolkit.org
> http://lists.einsteintoolkit.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.einsteintoolkit.org/pipermail/users/attachments/20250630/33be5811/attachment-0001.htm>


More information about the Users mailing list