How to Build PhasicFlow

You can build PhasicFlow for different hardware, CPU or GPU, depending on the computer. You can have a single build or oven multiple builds on a machine. Here you learn how to build PhasicFlow in various modes of execution.

Required packages

PhasicFlow uses Kokkos as the backend for parallelization. So, you need to have a copy of the source code in the machine alongside PhasicFlow. The make system is adjusted in a way so you do not need to separately compile Kokkos and the required source code files from Kokkos are compiled alongside the PhasicFlow.

g++ (C++ compiler)

The code is tested with g++ (gnu C++ compiler). The standard version of g++ on Ubuntu 18.04 LTS or upper is sufficient for compiling. If it is not installed on your operating system, enter the following command:

sudo apt install g++

CMake

You also need to have CMake-3.22 or higher installed on your computer.

tbb (2020.1-2 or higher)

For Ubuntu 20.04 LTS or higher versions, you can install tbb using official repository apt. For now, some parallel algorithms on host side rely on tbb parallel library (C++ parallel backend). Use e following commands to install it:

sudo apt update
sudo apt install libtbb-dev

If you are compiling on Ubuntu-18.04 LTS, you need to enter the following commands to get the right version (2020.1-2 or higher) of tbb:

wget "http://archive.ubuntu.com/ubuntu/pool/universe/t/tbb/libtbb2_2020.1-2_amd64.deb"
sudo dpkg --install libtbb2_2020.1-2_amd64.deb
wget "http://archive.ubuntu.com/ubuntu/pool/universe/t/tbb/libtbb-dev_2020.1-2_amd64.deb"
sudo dpkg --install libtbb-dev_2020.1-2_amd64.deb

How to build?

PhasicFlow uses CMake as the build system. you need to have CMake-3.22 or higher installed on your machine. you can either use the command line for setting-up the build system, or use cmake-gui to setup your build system through the GUI.

Step 1: Package check

Make sure you have installed all the required packages on your computer (tbb and CMake and compilers).

note: You must have git installed on your computer. If it is not installed, enter the following command:

sudo apt install git

Step 2: Cloning Kokkos

It is assumed that Kokkos source is located in the home folder of your OS. Clone the current version of Kokkos into your home folder:

cd ~
mkdir Kokkos
cd Kokkos
git clone https://github.com/kokkos/kokkos.git

or simply download and extract the source code of Kokkos in ~/Kokkos folder. In the end, the top level CMakeLists.txt file should be located in ~/Kokkos/kokkos folder.

Step 3: Cloning PhasicFlow

Create the PhasicFlow folder in your home folder and then clone the source code into that folder:

cd ~
mkdir PhasicFlow
cd PhasicFlow
git clone https://github.com/PhasicFlow/phasicFlow.git

Step 4: Environmental variables

In the terminal enter the following command:

gedit ~/.bashrc

and add the following line to the end of the file, save and close it.

source $HOME/PhasicFlow/phasicFlow/cmake/bashrc

close the terminal.

Step 5: Building PhasicFlow

Follow one of the followings to build PhasicFlow for one mode of execution.

Serial build for CPU

Open a new terminal and enter the following commands:

cd ~/PhasicFlow/phasicFlow
mkdir build
cd build
cmake ../ -DpFlow_Build_Serial=On
make install

For faster builds, use make install -j. This will use all the CPU cores on your computer for building.

OpenMP build for CPU

cd ~/PhasicFlow/phasicFlow
mkdir build
cd build
cmake ../ -DpFlow_Build_OpenMP=On
make install

GPU build for parallel execution on CUDA-enabled GPUs

cd ~/PhasicFlow/phasicFlow
mkdir build
cd build
cmake ../ -DpFlow_Build_Cuda=On
make install

After building, bin, include, and lib folders will be created in ~/PhasicFlow/phasicFlow/ folder. Now you are ready to use PhasicFlow.

Step 6: Testing

In the current terminal or a new terminal enter the following command: checkPhasicFlow This command shows the host and device environments and software version. If PhasicFlow was build correctly, you would get the following output:

Initializing host/device execution spaces . . . 
  Host execution space is Serial
  Device execution space is Cuda

  You are using phasicFlow v-0.1 (copyright(C): www.cemf.ir)

Finalizing host/device execution space ....