Platforms platform#
Platform plugins: environment setup and job submission for HPC systems.
A platform plugin abstracts the hardware- and scheduler-specific details of a computing environment so that the rest of pyCIF can submit and monitor jobs in a portable way. Every platform plugin must expose three callables:
ini_data(plugin)— load environment modules and set model-specific compiler/library paths on the current node.submit_job(self, exe, job_file)— write and submit a batch job script for the given executable command; return the job identifier.check_jobs(self, list_jobs)— returnTruewhen all listed jobs have finished.
Available platforms#
Available Platforms platform#
The following platforms are implemented in pyCIF so far:
- Bologna’s ECMWF cluster
ECMWF/ecs - CSC-Puhti’s FMI cluster
FMI/Puhti - CSCS - Swiss National Supercomputing Centre
EMPA/daint - Centre de Calcul Recherche et Technologie (AMD/rome)
TGCC-CCRT/AMD - Centre de Calcul Recherche et Technologie (skylake)
TGCC-CCRT/std - Centre de Calcul Recherche et Technologie with NVIDIA environnement
TGCC-CCRT/nvidia - Docker container for pycif
docker/cif - Docker container for pycif with nvidia HPC environment
docker/nvidia - ESPRI Spirit/SpiritX cluster with NVIDIA environment
ESPRI/spirit-nvidia - IDRIS Jean-Zay cluster
JEAN-ZAY/std - LAERO’s cluster
LAERO/nuwa - LSCE’s cluster
LSCE/obelix - LSCE’s cluster with NVIDIA environnement
LSCE/obelix-nvidia