Overview

Some software may not be available as a module or through Conda. In these cases, you can compile and install software from source. Doing so allows you to customize the installation, optimize for specific nodes, and control versions.

Best Practices

• Install in your home or project space: Avoid using scratch directories because they are temporary.
• Check node architecture: Some software may benefit from optimization flags for specific nodes (Haswell, Skylake, Milan, Xeon Max).
• Use modules for dependencies: Load any necessary compilers or libraries with module load before building.
• Document your install: Keep a log of commands, configuration flags, and installed locations for reproducibility.

Typical Installation Steps

# Unpack source code
tar -xvf software-version.tar.gz
cd software-version

# Configure build options
./configure --prefix=/gpfs/home/username/software/software-version

# Compile
make -j 8

# Install to the chosen directory
make install
    

Note: Adjust the --prefix path to your home or project space. Use -j N in make to match the number of cores you want to use for compilation.

After Installation

export PATH=/gpfs/home/username/software/software-version/bin:$PATH
export LD_LIBRARY_PATH=/gpfs/home/username/software/software-version/lib:$LD_LIBRARY_PATH
      

• Add the installation directory to your PATH or environment variables as needed:
• Consider creating a modulefile if multiple users will need access to the software.