I7 Logo
Chair for Foundations of Software Reliability and Theoretical Computer Science
Informatik Logo TUM Logo
A GPU-based PG solver

Overview

The GPUPG package contains GPU-based implementations of several algorithms for solving parity games. Currently the package only contains a solution for MS Visual Studio 2010. For the CUDA versions please install the Nvidia CUDA toolkit (tested v5.0). For the OpenCL versions of the solvers you will need depending on your system either the Nvidia CUDA toolkit (tested v5.0, only GPU) or the AMD APP SDK (tested v2.7, GPU and CPU) or the Intel OpenCL SDK (tested beta 2013, only CPU). Corresponding project files are included in the solution. (So far only Win32 targets are defined, but x64 should work, too.) We currently use the input format of PGSolver for parity game arenas. To run the solvers, simply pass the filename of the arena to the respective solvers.

Source code and benchmark arenas

Contact

  • Philipp Hoffmann (Email: ph DOT hoffmann AT tum DOT de)
  • Michael Luttenberger (Email: luttenbe AT model DOT in DOT tum DOT de)