Re: request C code which translates source into PDG

cliffc@rice.edu (Cliff Click)
Wed, 14 Apr 1993 23:02:47 GMT

          From comp.compilers

Related articles
request C code which translates source into PDG wjackson@pinnacle.cerc.wvu.edu (1993-04-13)
Re: request C code which translates source into PDG cliffc@rice.edu (1993-04-14)
Re: request C code which translates source into PDG paco@legia.cs.rice.edu (1993-04-15)
Re: request C code which translates source into PDG krishna@cs.unm.edu (1993-04-15)
| List of all articles for this month |

Newsgroups: comp.compilers
From: cliffc@rice.edu (Cliff Click)
Keywords: tools, analysis, optimize
Organization: Center for Research on Parallel Computations
References: 93-04-047
Date: Wed, 14 Apr 1993 23:02:47 GMT

wjackson@pinnacle.cerc.wvu.edu writes:
> Has anyone implemented algorithms which convert a source program into a
> Program Dependence Graph? We are currently working to implement such
> algorithms as outlined in the 1987 paper "The Program Dependence Graph and
> Its Use in Optimization" by Ferrante, Ottenstein, and Warren.


I've implemented conversion of a low-level 3-address intermediate
representation to something similar to a PDG. I do analysis and
optimizations on this form, then produce something close to assembly out.


Paul Havlok (paco@cs.rice.edu) has implemented conversion of Fortran
(really an AST of the Fortran program) to something similar to a PDG.
Paul does analysis and optimizations, then outputs readable Fortran.


Our forms are more like each other than they are like a PDG. We both have
fewer restrictions on the scheduling of computations than the PDG does,
and use this freedom to do better analysis.


Because of our different purposes the kinds and extents of our analyses
differ, but the basic graph formats are very similar. We both found this
kind of IR is very easy to analyze and optimize.


I can send you C++ code for what I do, but its probably less useful than a
careful description of how you would go about doing it.


Cliff
--
cliffc@cs.rice.edu
Massively Scalar Compiler Group
--


Post a followup to this message

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