Gysela Mini App I/O

A minimal application demonstrating GYSELA I/O operations and testing the CPU performance scaling for 5D particle distribution functions.

This repository contains mini apps that allow easy coupling with GyselaX++.

Installing

git clone git@github.com:gyselax/gysela-mini-app_io.git
cd gysela-mini-app_io
git submodule update --init --recursive

Sourcing the environment

source external/gyselalibxx/toolchains/<MACHINE>/environment.sh

For more details see Gyselalib++ environment toolchains.

Building

cmake -S . -B build -DCMAKE_TOOLCHAIN_FILE=external/gyselalibxx/toolchains/<MACHINE>/toolchain.cmake
cmake --build build -j 4

if you want to use Python insitu-diagnostics set additionally the PYTHONPATH:

export PYTHONPATH=/home/PK282018/develop/gysela-mini-app_io/python:$PYTHONPATH

Running

mpirun -n <nprocs> ./build/apps/gys_io <config.yaml> <pdi_config.yaml>

Example

mpirun -n 4 ./build/apps/gys_io apps/gys_io.yaml apps/pdi_default.yaml

Note that nprocs>1 works currently only with the version "mpi_transpose" switched on in the apps/gys_io.yaml yaml file.

Configuration

Edit apps/gys_io.yaml to configure:

• Mesh: Grid sizes and ranges for toroidal coordinates (Tor1, Tor2, Tor3) and velocity space (Vpar, Mu)
• Species: Number of species, charges, masses
• Application version: "mpi_transpose" or "in-situ-diagnostic"

Output Files

• fdistribu_5D_output.h5: Distribution function and mesh coordinates
• cpu_time.h5: CPU timing statistics (initialisation, transpose, GPU↔CPU transfer, I/O)
• fluid_moments.h5: Fluid moments (density, mean velocity, temperature)

Overview

This mini application:

• Initialises a 5D particle distribution function (species × toroidal coordinates × velocity space)
• Writes the distribution function and mesh coordinates to HDF5 files
• Computes fluid moments (density, mean velocity, temperature) via C++ integration or in-situ Python computation
• Measures and saves CPU timing statistics