Getting Intel MKL Installed on Ubuntu 18.04

So, I totally get how confusing it can be to get Intel MKL up and running on your Ubuntu set up. Here’s a little step-by-step that might help clear things up!

1. Downloading MKL

First, go to the Intel oneAPI MKL page. You can try the free version, which is called “oneAPI”. Typically, you’ll want to download the Linux version. There’s usually a .tar.gz file that’s the right choice for Ubuntu.

2. Installing MKL

Once you’ve downloaded the file, you can extract it using the terminal:

tar -xzf l_mkl_*.tar.gz

Now, navigate to the extracted folder and run the installation script:

cd l_mkl_*/install
sudo ./install.sh

Just follow the prompts during installation. You can usually go with the default options unless you have something specific in mind.

3. Setting Environment Variables

After installing, you will need to set up your environment variables. Open your .bashrc file:

nano ~/.bashrc

Add the following lines at the end (make sure to replace {path-to-mkl} with your actual installation path):

export LD_LIBRARY_PATH={path-to-mkl}/lib:$LD_LIBRARY_PATH
export MKLROOT={path-to-mkl}

Then run this to apply the changes:

source ~/.bashrc

4. Testing the Installation

You can check if it’s installed correctly by compiling and running a simple example. Create a file called test_mkl.c and add this code:

#include 
#include "mkl.h"

int main() {
    double result = cblas_ddot(3, (double[]){1, 2, 3}, 1, (double[]){4, 5, 6}, 1);
    printf("Dot product: %f\n", result);
    return 0;
}

To compile it, use:

gcc test_mkl.c -o test_mkl -I$MKLROOT/include -L$MKLROOT/lib -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -lpthread -lm

If it compiles without error, run it with:

./test_mkl

If you see a dot product result printed out, congrats! MKL is up and running!

5. Troubleshooting

If you run into errors, double-check your paths and ensure all dependencies are installed. You might need to install additional packages using:

sudo apt-get install build-essential

Forum posts and places like Stack Overflow are great for troubleshooting specific error messages.

Final Thoughts

It can feel overwhelming, but once you get through it, using MKL is really speedy! Good luck, and just take it step by step. You got this!