Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site utcsrgv.UUCP
Path: utzoo!utcsrgv!phyllis
From: phyllis@utcsrgv.UUCP (Phyllis Eve Bregman)
Newsgroups: ont.events
Subject: UofT DCS Seminar Schedule
Message-ID: <1517@utcsrgv.UUCP>
Date: Mon, 13-Jun-83 10:59:43 EDT
Article-I.D.: utcsrgv.1517
Posted: Mon Jun 13 10:59:43 1983
Date-Received: Mon, 13-Jun-83 14:12:01 EDT
Organization: CSRG, University of Toronto
Lines: 52


	UofT Department of Computer Science Seminar Schedule for
		     the week of June 13th, 1983


Thursday, June 16th, 3:00 P.M., SF1101:  Dr. Philip Rabinowitz, The
  Weizmann Institute of Science, Rehovot, Israel:  "Software for
  multiple numerical integration".

Thursday, June 16th, 4:00 P.M., SF1105:  James R. Cordy, Computer
  Systems Research Group, University of Toronto, "An orthogonal
  model for code generation".


			     ABSTRACT

     Code generation is the part of programming language  com-
     pilation  which is concerned with choosing the actual imple-
     mentation of programming language  constructs  in  terms  of
     computer  hardware.  We can characterize the code generation
     problem as a mapping from the abstract,  mathematical  world
     of  programming languages to the concrete, discrete world of
     computing machines.  Objects in each of these worlds can  be
     expressed  in  terms  of two fundamental concepts: operators
     and operands.
     
     Viewed in this way, every existing systematic approach to
     code generation is based on the same model.  This tradition-
     al model begins by decomposing  abstract  operand  structure
     into sequences of abstract operators. Machine implementation
     is then accomplished as a single complex process which tends
     to  obscure  the  mapping between language objects and their
     machine representation.
     
     We propose a new model,  which  we  call  the  orthogonal
     model, which separates the implementation of abstract opera-
     tors and abstract operands into two essentially  independent
     parts.   In the orthogonal model, abstract operand structure
     is left intact until implementation of abstract operators is
     complete.  It then remains to implement abstract operands as
     a separate step.
     
     We present a  code  generator  structure  based  on  this
     model,  which is simpler and more easily understood than ex-
     isting structures.  We present fundamental algorithms of the
     orthogonal  code generator and show how these algorithms can
     be easily parameterized across a large class of target  com-
     puters.

Friday, June 17th, 3:00 P.M., SF1105:  Professor Pavol Hell, Department
  of Computer Science & Mathematics, Simon Fraser University:  Generalized
  matchings in graphs".