Parallization of Stellar Atmosphere Codes

Parallel computing has turned out to be the enabling technology to solve complex physical systems. However, the transition from shared memory, vector computers to massively parallel, distributed memory systems and, recently, to hybrid systems poses new challenges to the scientist. We want to present a cook-book (with a very strong, personal bias) based on our experience with parallization of our existing codes. Some of the general tools and communication libraries are discussed. Our approach includes a mixture of algorithm, domain and physical module based parallization. The advantages, scalability and limitations of each are discussed at some examples. We want show that it becomes easier to write parallel code with increasing complexity of the physical problem making stellar atmosphere codes beyond the classical assumptions very suitable.