info sought on parallel mapping (compiler) problems

Hello
              I am interested in compiler problems, more specifically mapping
problems, on heterogeneous processing systems. So far what I have come
across in some literature in the mapping field are mainly for homogeneous
multiprocessor and in graphical approaches (since the general mapping
problems are NP-complete in strong sense, most I found using heuristic
graph approach).
Here are ideas on the problems extracted from some literature.
1. The problems start from finding a suitable program abstraction
(task graph) which can be derived from users' program. More methods may be
involved such as task clustering, partitioning and etc. These methods
require information on target architectures.
2. Finding near-optimal algorithms for mapping the program
abstraction onto a target machine. Some use architecture abstraction
(architecture graph).


Any pointers , suggestions and comments (bibliography) on the
topics are greatly appreciative. I'll summarize the result.
--
Regards,


Chokchai Leangsuksun
Graduate Assistant
Department of Mathematics and Computer Science
Kent State University
Kent Ohio 44242
e-mail: chokchai@mcs.kent.edu
--

Return to the comp.compilers page.
Search the comp.compilers archives again.