NeoArealize

A model of reaction-diffusion pattern formation in neocortex.

Pre-requisites:

Build and install morphologica from:

https://github.com/ABRG-Models/morphologica

Make sure these packages are installed (Debian/Ubuntu example):

sudo apt install python python-numpy xterm

Build and install jsoncpp (in a directory '~/src', just for example, you can build it wherever suits):

mkdir -p ~/src
cd ~/src
git clone https://github.com/open-source-parsers/jsoncpp.git
cd jsoncpp
mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=/usr/local -DBUILD_SHARED_LIBS=YES ..
make
sudo make install

Now you can build NeoArealize:

cd NeoArealize
mkdir build
cd build
cmake ..
make
cd ..

To run:

./build/sim/james1 ./configs/c1.json

The program reads parameters from c1.json and writes results into ./logs/

The program is multi-threaded (using OpenMP pragmas). To get the best performance it's usually necessary to experiment. Use a computation only program (like james1c) and set different values for OMP_NUM_THREADS. For example:

export OMP_NUM_THREADS=10 && time ./build/sim/james1c ./configs/c1.json >/dev/null

Or use the findfastest.sh script in misc/.

On an Intel i9 7980XE with 18 cores, 13 seems to be fastest, but there's not much difference for the range 6 cores to 18 cores.

On an AMD Threadripper 2990WX with 32 cores, about 20 seems to be fastest. This CPU is slower for these reaction diffusion equations than the 7980XE. (2 seconds for the Intel, 3 seconds for the AMD).

ABRG-Models/NeoArealize

NeoArealize