home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.parallel
- Path: sparky!uunet!gatech!hubcap!fpst
- From: mccurley@cs.sandia.gov (Kevin S. McCurley)
- Subject: Re: Description of C*
- Message-ID: <1992Jul27.140906.25891@hubcap.clemson.edu>
- Apparently-To: comp-parallel@ucbvax.berkeley.edu
- Sender: usenet@cs.sandia.gov (Another name for news)
- Organization: Sandia National Laboratories, Albuquerque, NM
- References: <1992Jul23.134937.5098@hubcap.clemson.edu>
- Date: Fri, 24 Jul 92 04:21:51 GMT
- Approved: parallel@hubcap.clemson.edu
- Lines: 45
-
- In article <1992Jul23.134937.5098@hubcap.clemson.edu> peteh@cs.rhbnc.ac.uk (PeterR.Hoare) writes:
- >Could anybody tell me where I can get a good reference to the new
- >version of C* which is being proposed as a ANSI standard parallel C.
-
- OH NOOOOOOO!
-
- Perhaps I have not kept up with what is happening with data parallel
- C, but this smells of a potential disaster. For three years I have
- been disgusted with the lack of any reasonably flexible model of
- data-parallel C. I am not an expert in computer languages, but I
- write considerable amount of code for parallel machines. For three
- years I kept thinking that if I just waited until a reasonable
- replacement came along for the CM-2, then maybe we would see some
- reasonable data-parallel version of C come along to replace C*.
- Numerous times I have thought about writing some application in C*,
- only to realize that it would be severely limited by the lack of
- parallel pointers. In my opinion adopting C* as the base for
- data-parallel C is ludicrous, since it is far too restrictive. C* was
- designed years ago for a machine whose processors were incapable of
- doing indirect addressing, and required every processor to have an
- identical memory map. It has some nice features, but it has a long
- way to go until it will be powerful and flexible enough to allow
- people to parallelize existing serial code in a natural way.
-
- I would hope that any new standard would get us beyond this, and look
- forward to what programmers will want to use in the future. In my
- opinion it is far too early to adopt such a standard, and the parallel
- computing community (including TMC) would be better served by letting
- data parallel C mature more before pushing for a standard. In the
- meantime there is a glaring need for more work to be done on this
- subject.
-
- I am simply a scientist who has longed for a data-parallel C language
- that would allow me to express my algorithms in a straightforward and
- efficient manner. I am not a language expert, so it is possible that
- I do not understand all of the issues involved. On the other hand,
- you don't have to be a racehorse breeder to recognize a horse with no
- teeth and a potbelly.
-
- Kevin McCurley
- Sandia National Laboratories
-
-
-
-
-