Experiments in Sustainable Software Practices for Future Architectures
This work addresses software sustainability for researchers and developers at national laboratories dealing with high-performance computing on emerging architectures, but it is incremental as it focuses on awareness-raising experiments rather than major breakthroughs.
The authors tackled the challenge of adapting large physics codes for new architectures like many-core, GPU, and Intel MIC by exploring sustainable software practices, reporting on two small experimental projects at LANL to raise awareness of these approaches.
In the process of rewriting large physics codes at Los Alamos National Laboratory to perform well on new architectures such as many-core, GPU, and Intel MIC, we have found a number of areas in which sustainable software practices can provide significant advantages. We describe several specific advantages of sustainable practices for future architectures, and report on two small experimental projects at LANL intended to raise awareness of new software practices and programming approaches for new architectures.