Dynamic Scheduling of Parallel Applications on Shared-Memory Multiprocessors Tight Cooperation between the Compiler

iii iv Current parallel shared-memory multiprocessors are complex machines, where a large number of architectural aspects (both of processors and memory) have to be simultaneously addressed in order to achieve high performance. The quick evolution of the parallel machines has been followed by the evolution of the parallel execution environments. Both user and kernel execution levels are the subject of current research works that try to achieve a good cooperation between them and with the hardware. These were the trends we were interested in, when we started the development of this thesis. The thesis proposes a parallel execution environment, where is possible to integrate new approaches both at user and kernel levels. The main goal is to achieve both efficient and effective support for multi-user parallel processing across a wide range of shared memory parallel computers, from small symmetric multiprocessors to high-end systems with globally shared address space