Discamus continentiam augere, luxuriam coercere
Home -> Research -> NB Collectives -> Kernels
NB Collectives
      CG Solver
    MPI Topologies
    MPI Datatypes
    Network Topologies
    Ethernet BTL eth
    Older Projects
  Full CV [pdf]


  Past Events

Application Kernels using LibNBC

This page lists several application kernels that have been implemented to demonstrate the optimization potential of LibNBC. The only kernel available so far is the parallel implementation of a three-dimensional Fast Fourier Transformation. The 1-dimensional transformation is optimized for communication not for cache-reuse. This means that the imlementation might be slower on standard machines than a different domain decomposition.

3d-FFT (pre-release) - (5.6 kb)

serving:© Torsten Hoefler