automatic exploitation of task parallelism from sequential

> Is it true that most researchers in parallel computing, consider
> automatic exploitation of parallelism from sequential programs, like
> C, too complicated ?


Mostly. The trouble is that C is particularly grotty in this respect
as it is very predicated on sequential execution and leaves few
avenues for algorithmic transformation. Fortran, particularly F90 is
a better vehicle as the method of implementation has not been overly
defined. Ada has some potential likewise. But most work tends to be
done on special purpose languages (though more often this turns out
to be a cut-down C or Fortran with all the awkward bits thrown away
and some new facilities). You should certainly survey the Fortran
work in this regard - primarily High Performance Fortran which is
mostly interested in distributing arrays over many processors. Sorry
I don't have a URL, but the HPF spec is available as a Technical
Report from Rice University and a lot of work has been done at the
University of Vienna (look for Fortran D as well).


Good luck!
----------------------------------------------------------------------
Dave Lloyd Email: Dave@occl-cam.demon.co.uk
Oxford and Cambridge Compilers Ltd Phone: (44) 1223 572074
55 Brampton Rd, Cambridge CB1 3HJ, UK
--

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