Effective use of multicore-based parallel computers for scientific computing

This thesis studies how the multi-core hardware architecture can be efficiently used for real-world scientific applications that arise from computational cardiology and computational geoscience. The investigation has been carried out from different angles: numerical algorithms, parallel programming and performance modeling and prediction. It is shown that high-performance implementations and optimizations must match both the underlying computations and the target parallel platform. Several good practices are summarized for parallel programming and performance analysis on the multi-core architecture, which can be of help to many other scientists