Programming with UG

The UG high-level task parallelization framework provides a systematic way to develop a massively parallel solver, that can run on large scale supercomputers.

1. develop and debug a stand-alone Solver, if the Solver does not exit.
2. develop and debug a shared-memory parallel solver, in which Solvers are managed by LoadCoordinator.
3. develop and debug a distributed-memory parallel solver, in which Solvers are managed by LoadCoordinator.

Note that, algorithmically, 2 and 3 work the same. Therefore, a massively parallel solver can be developed on a PC.

• Classes hierarchy and source code directory organization
• Coding style guidelines
• Debugging