Dense and CSR Sparse matrix library in C++ leveraging Intel oneMKL (CBLAS)

What this project does

In this project of ours we are developing a high-perfomance Dense and CSR Sparse matrix library in modern C++ (icpx), leveraging Intel oneMKL (CBLAS + MKL Sparse BLAS) for numerical computing and HPC applications.

How to build & compile the project

The following build and compile instructions are for Linux (we will assume that you are using Ubuntu/Debian like distros). If you use Windows you may encounter problems.

Requirements

Make sure you have all of the following installed before reading the Setup Instructions section.

• Git:

  https://git-scm.com/install/linux

  or

  sudo apt-get install git

• CMake:

  sudo apt-get install cmake

• Intel oneAPI Base Toolkit:

  https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit-download.html

• VS Code (or you IDE of preference but we will assume you use VS Code below)

  You usually can install it directly from the App Center or from here

Setup Instructions

1. First you will need to clone the repository to your desired destination.

   Go to the directory you want to clone the project:

   cd <path_to_your_dir>

   Clone the repo:

   git clone https://github.com/VictorMeimariss/Cpp-Project.git

   Then get inside the project root folder: cd Cpp-Project

2. Next, you need to source/setup the oneAPI variables:

   source /opt/intel/oneapi/setvars.sh

   The path above is usually the default, it may vary if you changed anything in the installation process

3. Open your IDE:

   For VS Code do:

   code .

Build & Linking Instructions

Now to run the CMake Configuration run the following command in the VScode terminal:

cmake -S . -B obj

this will create a directory called /obj where all the .o files will go.

Next, Build and Link everything with the following command:

cmake --build obj

Run a test (optional)

Now to confirm that everything went well you can run one of the test functions .cpp files (all the executables are inside /bin)

./bin/test_CG_dsolve 

it doesn't need to be test_CG_dsolve, this was just used as an example.