We have two ongoing Fortran based compiler efforts to target irregular
problems. Using the CHAOS library and the Syracuse Fortran 90D compiler we have developed
a prototype distributed memory compiler able to generate efficient code for templates
extracted from adaptive problems. By making use of the Rice D System, we have developed
loop slicing methods capable of dealing with unstructured routines with multiple levels of
distributed indirection. Finally, we have also applied CHAOS directly to parallelize a
number of full adaptive applications codes.
We are also a part of the Parallel Compiler Runtime
Consortium (PCRC), a project to produce the runtime infrastructure to support several
data parallel language compilers.
