SIAM_2017.pdf

Software complexity has become a real barrier that impedes scientific software development, such as adding new features and functions, validating domain knowledge incorporated in the software systems, as well as redesigning and refactoring code for emerging computational platforms (i.e., exascale computers). In this poster, we analyze several environmental software systems and demonstrate software design and refactoring challenges encountered by scientific community.  We also summarize practices that leverage compiler technologies for better program understanding and enhanced software productivity. These technologies have been used to (1) understand existing scientific models, (2) modularize complex code, (3) provide instrumentation mechanisms to improve code portability, (4) generate functional testing unit for key software modules, and (5) add new features for model validation. These activities are supported by the Accelerated Climate Modeling for Energy (ACME) project and the Interoperable Design of Extreme-scale Application Software (IDEAS) project.  We believe our experience can benefit broader scientific communities that are facing the challenge of complex code.