Automatic Parallelization: New Approaches to Code Generation, Data Distribution, and Performance Prediction

This work contains 11 contributions which deal with true automatic parallelization, and the focus is on automatic methods. Some of the questions under discussion are: up to which degree is automatic parallelization for DMS possible today? What are currently the most important problems for automatic parallelization and which new ideas are there? In which cases can knowledge-based methods help? Are there promising methods for automatic data distribution and redistribution? Why is performance prediction problematic?