Speaker
Description
The scientific computing community is suffering from a lack of good development tool that can handle well the unique problems of coding for high performance computing. It is much more difficult for domain experts to parallelize inherited serial codes written in FORTRAN which are very common in CSE research field. An automatic parallel programming IDE is developed for rapid development of large-scale numerical simulations. A domain specific graphical user interface is designed ot facilitate the algorithm design and parallel program construction based on current serial codes. A debugging approach based on the application development mechanism is built in to locate parallel errors automatically. Real applications demonstrate that is helpful on developing complex numerical applications and enabling increased software productivity greatly.
