OPTIMIX home page

OPTIMIX is an optimizer generator. It can be used to generate program analyses and transformations. Its input language is based on DATALOG and graph rewriting. Especially two new classes of graph rewrite systems are used: edge addition rewrite systems (EARS) and stratified graph rewrite systems (stratified GRS).

OPTIMIX can be used in the context of the CoSy compiler framework, with the Cocktail toolbox, in a standalone mode, or in Java embedded mode. For more information, contact the author or the company that sells CoSy (ACE).

Release 2.5 in tgz-format (4.9 MB, 98/11/13, with Sun4 and i686 executables)

Optimix can eat and produce Java. OPTIMIX graph rewrite systems are embedded in Java classes as methods. Java Vectors and Arrays are recognized. Java set functors can be defined. Path expressions can abbreviate Datalog chain rules. Set-based IN-expressions facilitate predicate specification (in the style of graph and path expression databases). If-then syntax for rules. Patterns in ()-brackets.

On-line release notes in html

Language manual 2.5 in ps.gz-format. [html-format]

Tutorial in ps.gz-format.

Release 2.5 (http protocol) (4.9MB, 98/09/24)

Release 2.5alpha in tgz-format. (2.1MB, 98/09/24) Language manual 2.5alpha in ps.gz-format.

Release 2.4alpha in tgz-format. (1.3MB, 97/8/11)

Release 2.4alpha in tar-format only (3.8 MB) Allows external functors (external graph libraries) and external intermediate representations (external node allocations). Foreign frontends and tools can be coupled to Optimix-generated code.

Release 2.3 in tgz-format. (97/4/29: bugfix release, language has not changed)

On-line release notes 2.2

Postscript language manual 2.2 (CoSy-8.0)

2.2: Choice-strata, termination check for rule groups.

2.1: Differences only for CoSy mode.

2.0: Standalone (AST) mode.

Up to release 2.0, OPTIMIX has been developed in the Esprit project COMPARE (No. 5399).

How to generate optimizers and their parts

In order to generate optimizer parts with OPTIMIX we propose the following procedure.

Define the data model of your application in Java, Cocktail-AST, or CoSy-fSDL i.e. define which parts of the knowledge you want to present should be objects and which should be graphs (relations).
Write down all preconditions for a transformation, perhaps in text. Starting from that, design of the data manipulation, i.e. formulate graph rewrite systems that compute and transform the graphs that were defined in the data model. Build graphs with edge addition rewrite systems (EARS), and transform them via general graph rewrite systems (GRS).
Think about the implementation of the graphs. Which algorithms does OPTIMIX generate for a problem and with which graph implementations do these run fastest? Exchange graph implementations (functor calls) accordingly. Tune graph rewrite systems by rule and procedure options to allow OPTIMIX to do better code generation.

Go to Uwe Assmann's home page

Last modified on Mon Oct 14 09:13:39 MET DST 1996 by Uwe Assmann.