ScalFMM is a scalable kernel independent fast multipole method. The design of the library separates the FMM-core (octree management, data structure, parallelism, ...) and the kernels that allow you to easily introduce new kernel. We provide as well as a kernel independent method based on interpolation (Chebyschev and equispaced grid points) for both scalar and tensorial kernels and different implementations for Poisson kernel (spherical expansion with rotation and BLAS optimizations). The library offers shared memory (OpenMP and StarPU), hybrid distributed/shared memory (MPI/OpenMP) parallelization and also on the runtime system (StarPU) for manycore architectures.
- improvements of the MPI version of ScalFMM
- ScalFMM on StarPu (runtime system) is available
- Add P2P on GPU
- Add ufmm (interpolation FMM based on uniform grid points)
- Fix bug in distribution generation