Die SpeedIT Tools Bibliothek stellt eine Reihe von beschleunigten Gleichungslösern für dünnbesette Matrizen oder Gleichungen zur Verfügung. Manifold acceleration, e.g. more than an order of magnitude, is achieved with a single reasonably priced NVIDIA Graphics Processing Unit (GPU) that supporst CUDA and proprietary advanced optimisation techniques.
SpeedIT is distributed in two versions: GPL-based and commercial. The first one, called SpeedIT Classic has been released for demonstration purposes and its performance has not been optimized, it supports also only single precision. New users are encouraged to download and test SpeedIT Classic in their programs to understand the API and installation procedure. Together with SpeedIT Plugin to OpenFoam® anyone can test this version also with OpenFOAM.
SpeedIT is a library of one of the fastest GPU implementations for Sparse Matrix-Vector Multiplication, Conjugate Gradient and BiConjugate Gradient Stabilized. In average it outperforms CUSP and CUSPARSE by following factors:
SpeedIT eXtreme has the same features as standard version but its internal kernel parameters are opitmized for a given GPU architecture. As such SpeedIT eXtreme takes the full advantage of your GPU card and therefore works fasters.
SpeedIT Classic implements classical algorithms and works in single precision only.