Parallel/High Performance Object-Oriented Scientific Computing 2008

摘要

Ever-increasing compute capability has enabled scientific programming to reach an unprecedented degree of sophistication and complexity. Complex algorithms, a wide range of hardware environments, and an increasing demand for software system modularity, portability and fault tolerance have shown that language-level abstraction must increase. At the same time, because the achievable capability of the highest-end machines defines the achievable limits of size and fidelity of scientific simulations, performance cannot be compromised.