In adaptive irregular problems the data arrays are accessed via indirection arrays, and data access patterns change during computation. Implementingsuch problems ondistributed memory machines requires support for dynamic data partitioning, e cient preprocessing and fast data migration. This research presents e cient runtime primitives for such problems. This new set of primitives is part of the CHAOS library. It subsumes the previous PARTI library which targeted only static irregular problems. To demonstrate the e cacy of the runtime support, two real adaptive irregular applications have been parallelized using CHAOS primitives: a molecular dynamics code (CHARMM) and a particle-in-cell code (DSMC). The paper also proposes extensions to Fortran D which can allow compilers to generate more e cient code for adaptive problems. These language extensions have been implemented in the Syracuse Fortran 90D/HPF prototype compiler. The performance of the compiler parallelized codes is compared w...
Shamik D. Sharma, Ravi Ponnusamy, Bongki Moon, Yua