1. Install prerequisites (You may contact the administrator to help you with this part)

    # already satisfied
CMake >= 3.10.2
MPI >= 3.0
Python >= 3.6
automake >=1.15

    # install requirements using root
sudo apt-get install openmpi-bin libopenmpi-dev libncurses5-dev libncursesw5-dev automake flex bison libncurses5-dev=5.0 libncursesw5-dev=5.0 cmake

  2. Install Nvidia-driver, CUDA and HPC-SDK

    <aside> 💬 Note: your HPC-SDK version should compatible with your CUDA version.

    </aside>

    NVIDIA HPC SDK Current Release Downloads

  3. Clone DeepDendrite repositoy

    git clone <https://github.com/pkuzyc/DeepDendrite.git>

  4. Compile NEURON simulator

    DeepDendrite provide a modified NEURON software, which you need to compile on your own environment.

    <aside> 💬 Note: Replace am__api_version='1.15' with am__api_version='1.16' in src/nrn_modified/configure for machines with latest version of Linux systems.

    </aside>

    cd src/nrn_modify
chmod 777 configure   # make it executable
./configure --prefix <abs_path/to/install/nrn_modified> --without-iv --with-paranrn --with-nrnpython=`which python` --without-memacs
make -j
make install

  5. Compile DeepDendrite:

    cd ../DeepDendrite

    Modify the install.sh file to load correct modulefile of PGI or NVHPC SDK, and set the install path and mod file path (all starts with "/path/to/your") in cmake command.

    <aside> 💬 On our A100 machine, you may replace install.sh with the followings:

    </aside>

    # install.sh file content
rm -rf build
mkdir build
cd build
export PATH=/opt/nvidia/hpc_sdk/Linux_x86_64/23.7/comm_libs/11.8/openmpi4/openmpi-4.1.5/bin:$PATH
export MANPATH=$MANPATH:/opt/nvidia/hpc_sdk/Linux_x86_64/23.7/comm_libs/mpi/man
export MANPATH=$MANPATH:/opt/nvidia/hpc_sdk/Linux_x86_64/23.7/compilers/man
export PATH=/opt/nvidia/hpc_sdk/Linux_x86_64/23.7/compilers/bin:$PATH
export PATH=/usr/local/cuda-11.8/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-11.8/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
# note: you'd better specify correct cuda version, "/usr/local/cuda/bin" might lead to unexpected error if there are multiple cuda toolkit with different versions

echo $(which mpicc)
echo $(which mpicxx)
export CC=mpicc
export CXX=mpicxx
cmake .. -DCMAKE_C_FLAGS:STRING="-lrt -g -O0 -mp -mno-abm" \\
    -DCMAKE_CXX_FLAGS:STRING="-lrt -std=c++11 -g -O0 -mp -mno-abm" \\
    -DCOMPILE_LIBRARY_TYPE=STATIC \\
		-DCMAKE_INSTALL_PREFIX=</path/to/install/DeepDendrite> \\ # example: /home/kchen/DeepDendrite/install/deepdendrite
    -DADDITIONAL_MECHPATH=</path/of/folder/for/your/nrn_mod_files> \\ # example: /home/kchen/DeepDendrite/src/all_mechanisms
    -DCUDA_HOST_COMPILER=`which gcc` \\
    -DCUDA_PROPAGATE_HOST_FLAGS=OFF \\
    -DENABLE_SELECTIVE_GPU_PROFILING=ON \\
    -DENABLE_OPENACC=ON \\
    -DAUTO_TEST_WITH_SLURM=OFF \\
    -DAUTO_TEST_WITH_MPIEXEC=OFF \\
    -DFUNCTIONAL_TESTS=OFF \\
    -DUNIT_TESTS=OFF
make -j
make install

    After finishing modifying install.sh, run it to complete installation.

    bash install.sh
# Note that everytime incoorperating new NEURON mod files, you need to reinstall DeepDendrite.

Run DeepDendrite

Running neuron simulation on DeepDendrite takes three steps.

Running on multiple GPU devices

Modify step 2 and 3. Here, we use 4 GPU devices as an example.