The GPU version of FARGO (GFARGO) is written in C and C for CUDA. It runs only on NVIDIA’s graphics cards. It should be regarded essentially as a "proof of concept". It corresponds to the standard, isothermal version of FARGO, although not all the functionnalities of the CPU version have been translated to CUDA. Namely, the following features are absent from the GPU version:
Gas accretion onto the planets
MPI communications. This version is therefore mono-GPU only.
-e command line flag (the one that triggers additional output for the EU comparison problem) has been removed.
The boundary conditions are exclusively the "damping boundary conditions" of the de Val-Borro’s et al (2006) paper.
The first release of GFARGO was exclusively available in single precision. This has some measurable effects, especially at large resolution. The download page also includes a newer release, compatible with FERMI architectures, which is a double precision version.
An additional feature is available for the case in which GFARGO is run on a graphics card connected to the display, which allows the user to see in real time how the fields evolve. The few snapshots below give an idea of this functionnality.