Releases: vortexlab-uclouvain/flups
Latest LGF kernels and some fixes
New LGF kernels and some bug fixes
- Fix some errors in the postprocessing of periodic directions
- Fix an error in the expression of the CHAT2 kernel
- Addition of new LGF kernels (LGF4 - LGF6 - LGF8 - MEHR4 - MEHR6) - Credits to J. Gabbard, MIT
FLUPS v2.0
In this release:
- Addition of support for node-centred data layout
- Addition of the Biot-Savart solver
- Improvement of the communication scheme
- Improvement of the distribution of unknowns across the MPI process
FLUPS v1.1
In this release:
- we extend the available kernels to higher order, and we introduce spectrally convergent kernels. Checkout the new
HEJ_0
kernel! - minor bug fixes
- improved documentation
FLUPS v1.0 - scalar and vector solver for 2-D and 3-D unbounded Poisson
Changelog
In this release, we introduce the resolution of a Poisson equation in 2/3D for a vector field, each component having a different set of BC's.
In order to store only once the Green's function in the fourier domain, we have to impose a common Fourier space to every component of the vector, starting from the 0-mode to the flip-flop mode.
To achieve this, regardless of the DCT/DST/DFT used, we correct the output of the DCT by adding the flip-flop mode and the output of the DST with a shift and the addition of the 0-mode.
This operation is local, hence scalable, and is done after/before every DCT/DST computation while in the correct topology.
To be achieved in the near future
- sample code for FORTRAN
- support of single-precision domains (float vs double).
- support of FORTRAN and C way of indexing
- support of vertex-centered meshes
Support for 2D and LGFs
We now support 2D problems, and Lattice Green Functions in 2D and 3D.
2D cases are obtained when specifying NONE boundary conditions in the 3rd direction.
Still to come until the 1.0 release:
- support for vector fields
- documentation on how to use the 2D solver
3-D version ready for production and scalability
Still to come in a later release:
- code adaptations for 2-D problems
- Lattice Green's Function kernels
- vector fields
- sample code for FORTRAN
- support for float
Known issues:
- only FORTRAN indexing convention is supported