Re: Different Strokes for Different Folks

purtilo@cs.umd.edu (Jim Purtilo) replied (in part):
> Some folks down at U Arizona produced a system for mixed language
> programming (called it MLP), written up in TSE and saw some distribution...


We here at The U. of Arizona are still involved in this work. ("We" being
the same advisor, different grad students :-) We have a system called
Schooner, an extension to the MLP work written up in TSE. We are
exploring problems of heterogeneity in scientific applications, and are
currently working with NASA Lewis Research Center as part of the NPSS
project. From a heterogeneity standpoint, the work is currently focused
on connecting different architectures to take advantage of the differing
programming models available. (We support language heterogeneity as well,
but are finding most work in the scientific community is in Fortran. That
doesn't leave a lot of use at present for our ability to support multiple
languages.) We are seeking not so much to speed-up the application, as to
allow applications to be connected in new ways. The idea being to provide
a seamless program where the user is not aware of the existence of
multiple codes within the program. Our goal is to simplify the task of
the application writer who wishes to join applications together to build
such seamless programs. In this context, heterogeneity is seen as an
advantage -- allowing the user to access the best machine for each
algorithm in the program without having to worry about the language
needed, data format conversions, etc.


Patrick Homer, patrick@cs.arizona.edu
PhD student
Dept. of Computer Science
The U. of Arizona
--

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