Installation

Prerequisites

MPI

HPCSparseArrays.jl requires an MPI implementation. When you install the package, Julia automatically provides MPI.jl with MPI_jll (bundled MPI implementation).

For HPC environments, you may want to configure MPI.jl to use your system's MPI installation. See the MPI.jl documentation for details.

MUMPS

The package uses MUMPS for sparse direct solves. MUMPS is typically available through your system's package manager or HPC module system.

Package Installation

From GitHub

using Pkg
Pkg.add(url="https://github.com/sloisel/HPCSparseArrays.jl")

Development Installation

git clone https://github.com/sloisel/HPCSparseArrays.jl
cd HPCSparseArrays.jl
julia --project -e 'using Pkg; Pkg.instantiate()'

Verification

Test your installation:

cd HPCSparseArrays.jl
julia --project -e 'using Pkg; Pkg.test()'

The test harness automatically spawns MPI processes for each test file.

Initialization Pattern

using MPI
MPI.Init()
using HPCSparseArrays
# Now you can use the package with BACKEND_CPU_MPI

Running MPI Programs

Create a script file (e.g., my_program.jl):

using MPI
MPI.Init()
using HPCSparseArrays
using SparseArrays

# Use the default MPI backend
backend = BACKEND_CPU_MPI

# Create distributed matrix
A = HPCSparseMatrix(sprandn(100, 100, 0.1) + 10I, backend)
b = HPCVector(randn(100), backend)

# Solve
x = A \ b

println(io0(), "Solution computed!")

Run with MPI:

mpiexec -n 4 julia --project my_program.jl

Troubleshooting

MPI Issues

If you see MPI-related errors, try rebuilding MPI.jl:

using Pkg; Pkg.build("MPI")

MUMPS Issues

If MUMPS fails to load, ensure it's properly installed on your system.

Next Steps

Once installed, proceed to the User Guide to learn how to use the package.