Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!crdgw1!uunet!europa.asd.contel.com!gatech!hubcap!fpst From: reinhard@rice.edu (Reinhard von Hanxleden) Newsgroups: comp.parallel Subject: Connection Machine Fortran Keywords: Connection Machine, Irregulare Problems, Vectorizers Message-ID: <1991May30.163634.17582@rice.edu> Date: 30 May 91 16:36:34 GMT Sender: news@rice.edu (News) Organization: Rice University Lines: 39 Approved: parallel@hubcap.clemson.edu Originator: reinhard@cimbria.rice.edu Dear netters, I would be grateful for comments/references on two issues regarding the Connection Machine. 1. IRREGULAR PROBLEMS (PIC codes, sparse matrices, _your favorite_, ...) - Which problems have been tackled? How did the implementation differ from equivalent MIMD implementations? - Have "classic" (MIMD) load balancing paradigms (e.g. recursive bisection) been applicable? - Any experiences in using indirection vectors to distribute a problem evenly? 2. GENERATING EFFICIENT CM-FORTRAN (CMF) FROM F77 / F90 - Who has experiences in tackling this with standard vectorizers? (One report I am aware of is by J. Myczkowski and R. Shapiro, trying the KAP vectorizer on 4 small samples; J. Mellor-Crummey has used PFC on a larger weather code.) - Have any vectorizers been modified for this task yet? - Any attempts to generate Layouts / Alignments, Foralls, or other CMF- specialities automatically from either F77 or F90? Have any major problems been identified yet? (Now, THIS should cause an avalanche, right?) Any piece of information will be appreciated, collected, and - if sufficient interest arises - be posted. Thanks in advance, Reinhard v. Hanxleden reinhard@rice.edu Department of Computer Science, Rice University P.O. Box 1892, Houston, TX 77251, (713) 527-4834 -- =========================== MODERATOR ============================== Steve Stevenson {steve,fpst}@hubcap.clemson.edu Department of Computer Science, comp.parallel Clemson University, Clemson, SC 29634-1906 (803)656-5880.mabell