• architectures
• cost models
• performance modelling
• shared memory
• structured parallel programming

Shared memory architectures are widely taking place. Following the structured parallel programming approach, a cost model is fundamental for performance portability and predictability. This thesis gives a contribution about cost models for multiprocessors and multi-cores taking into account important characteristics of new generation shared memory architectures, e.g. hierarchical shared memory, and with particular focus on the impact of the parallel application.