RFI: Compilers for supercomputers and parallel computers.

mnh@uts.amdahl.com (Mark Haynie)
3 Aug 89 21:00:48 GMT

          From comp.compilers

Related articles
RFI: Compilers for supercomputers and parallel computers. mnh@uts.amdahl.com (1989-08-03)
| List of all articles for this month |

From: mnh@uts.amdahl.com (Mark Haynie)
Newsgroups: comp.parallel,comp.compilers
Date: 3 Aug 89 21:00:48 GMT
References: <6039@hubcap.clemson.edu>
Organization: Amdahl Corporation, Sunnyvale CA

I'm looking for a good set of references regarding the most critical and
pressing issues with two areas in supercomputing: 1) supercomputing compiling
and 2) programming environments for parallel applications.

I am trying to come up the learning curve on the present state of the art with
regards to supercomputing. I've read quite a number of journal articles from
proceedings, trade journals, and textbooks. I have actually done some
mathematical programming and machine design on MIMD processors (mostly IBM) but
I have no practical experience with compilers or supercomputer design, however,
I understand the principles well enough. I'm really looking for a more solid
technical base of understanding.

I realize I may be asking for more than I expect, but what I'd like are
recommendations from the working community on a "reasonable enough" set of
references (mostly papers and textbooks right now) to the most significant
issues facing supercomputing environments. My suspicion is that many of these
issues today are more in software (i.e., compiler and programming environments)
than in hardware (there are quite a number of new machine architectures and
designs out there in both academia and industry).

In the area of supercomputing compiling, I'm interested in two things:

1) A good survey or comprehensive tutorial on the problem of program
concurrentization (versus vectorization),
2) What are the theoretical and practical challenges in dependence analysis and
program transformations for concurrentization?
3) Are there other theoretical or practical issues with supercomputer compiling
that are critical?

In the area of programming environments for parallel applications, I'm
interested in getting some feedback from people who write algorithms and
implementation them as programs on what they would expect from a supercomputer
programming and debugging environment.

1) What are the most pressing and critical issues with the usability of
supercomputer systems (languages, compiler support, program development and
debugging environment).
2) Is anyone making headway or doing the job right today? What's holding
progress up?

I'd appreciate people sending responses to me at this mail address. If there is
an appropriate newsgroup, I'd appreciate someone telling me where it is
(comp.arch and comp.compiler don't seem to particularly focus on
supercomputing/parallel processing issues).

-- Jim Garcia
      reply to sun!amdahl!mnh

Post a followup to this message

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