PetaVision is an open source, object oriented neural simulation toolbox optimized for high-performance multi-core, multi-node computer architectures.

Quick instructions for installing PetaVision and running the system tests:

Required dependencies:

• gcc/g++ (https://gcc.gnu.org/).
• bison (https://www.gnu.org/software/bison/).
• flex (https://www.gnu.org/software/flex/).
• cmake (http://www.cmake.org)

Optional (but strongly suggested) dependencies:

• OpenMPI (http://www.open-mpi.org).
  • You'll need mpicc, mpic++, and mpiexec.
• OpenMP (http://openmp.org/wp/)
• CUDA (https://developer.nvidia.com/cuda-downloads)
• cuDNN (https://developer.nvidia.com/cudnn)

Suggested additional tools:

• lua (http://www.lua.org/)
  • For designing parameter files for complex networks.
• octave (https://www.gnu.org/software/octave/)
  • To read/analyze data from OpenPV.
• python (https://www.python.org/)
  • To read/analyze data from OpenPV.
• mermaid (http://knsv.github.io/mermaid/)
  • For generating graphical drawings of networks from parameter files.

Building:

git clone http://github.com/PetaVision/OpenPV.git
mkdir build
cd build
cmake ../OpenPV
make

Other build options:

# Debug build (Release is the default)
cmake -DCMAKE_BUILD_TYPE:String=Debug ../OpenPV

# If CUDA is installed but you don't want CUDA support
cmake -DPV_USE_CUDA:Bool=OFF ../OpenPV

# To specify a particular CUDA architecture
cmake -DPV_CUDA_ARCHITECTURE:String=[arch]
# [arch] can be a compute capability e.g. 7.0, an architecture name e.g. Volta,
# "Auto" to detect and use the compute capability of the current GPU,
# or left blank to use the default choice for the CUDA version

# Build with clang address sanitization
cmake -DPV_ADDRESS_SANITIZE:Bool=ON ../OpenPV

Running the system tests:

cd tests
ctest

Our webpage is http://petavision.github.io/.

More detailed documentation is available at http://petavision.github.io/doxygen.

For general questions and discussion, post to our Gitter page: https://gitter.im/PetaVision/OpenPV