Numerical solvers

1 Multifrontal sparse symmetric linear solver

Key characteristics:

• Designed for: sparse, symmetric, indefinite systems.
• Efficient for sparse KKT-systems from constrained optimization, saddle point systems and similar from DAEs Differential Algebraic Equations and similar.
• Single-core implementation. Thread-based paralellization planned.

1.1 C-API

• Download: liblindefsol.so lindefsol.h
• Tested on Ubuntu 20.04, uses minimal system calls, should work on almost any modern Linux distribution.

2 Nonlinear constrained optimization solver compiler

Based on a nonlinear interior-point penalty method with the multifrontal sparse symmetric linear solver as its main workhorse.

Key characteristics: * Optimized for fast solutions to medium sized problems, eg. real-time optimal control.

Following workflow:

• Create nonlinear model description using a .json file.
• Compiler takes as input the .json file, and outputs an optimized nonlinear solver for the given problem.

• The optimized linear solver is output as a C-library, with a standard C-interface.

• Download: liboptimizer.so optimizer.h

• Tested on Ubuntu 20.04, uses minimal system calls, should work on almost any modern Linux distribution.