cpmech / laclib

Linear Algebra C Library

Simple Linear Algebra Library

Laclib implements basic linear algebra functions, wrapping OpenBLAS or Intel MKL. It also wraps MUMPS or Intel Direct Sparse Solver (DSS, e.g., PARDISO) for the solution of large linear systems with sparse matrices.

Examples

The following code shows how to solve a linear system with a (large) sparse matrix:

auto coo = read_matrix_market("my-matrix.mtx");
auto options = MumpsOptions::make_new(is_symmetric(coo->layout));
auto solver = SolverMumps::make_new(options);
auto rhs = vector<double>(coo->n, 1.0);
auto x = vector<double>(coo->n, 0.0);
solver->analyze_and_factorize(coo);
solver->solve(x, rhs);

See ex_solver_mumps.cpp for a complete code.

Use laclib in your project

After installation (explained below), you may clone https://github.com/cpmech/use-laclib to bootstrap a project using laclib.

Docker image

The docker image is convenient with Visual Code remote development tools (but containerized; "not really remote").

./build-docker-image.bash

Ubuntu/Linux 22.04+

First, follow the procedures from https://github.com/cpmech/script-install-mumps

Then, execute

./install-deps.bash
./install.bash

which will compile and install the header files in /usr/local/include/laclib and the library files in /usr/local/lib/laclib.

Directory structure

├── benchmarks   # tests for computational performance
├── data         # auxiliary data for unit tests
├── examples     # examples of usage
├── src          # this is the "main library"
│   ├── analysis # analysis, stats, and reporting
│   ├── check    # functions for unit/integration tests
│   ├── linalg   # basic linear algebra functions
│   ├── sparse   # sparse matrix and solvers tools
│   └── util     # some utilities for file/string manipulations
└── zscripts     # auxiliary "internal" bash scripts

We suggest looking at the sparse directory first.

Intel MKL Installation

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

Installing MKL on Linux

Select Linux → APT Package Manager and follow the instructions, or use the script below:

bash ./zscripts/install-intel-mkl-linux.bash

Installing MKL on Windows

Select Windows → Online Installer and follow the instructions.