Contents
 

• About Ursa Minor
• Objectives
• Overview
• Functionality
• Future Development
• About the Author
• Online User's manual
• Developers' Guide (internal use only)

The Ursa Minor (User Responsive System for the Analysis, Manipulation, and Instrumentation of New Optimization Research) project is an on-going effort to develop a parallel programming environment. The tool is designed to help understand the structure of a program, visualize, manipulate, and compare performance results under different environments, identify potential parallelism, and store the information for later use. It features an interactive user interface, and was written in 20,000 lines of Java code. The second generation of Ursa Minor (Ursa Minor 99) has been released and reached version 1.51.

The tool is closely related to the Polaris compiler infrastructure. Polaris, as a compiler, includes advanced capabilities for array privatization, symbolic and non-linear data dependence testing, idiom recognition, interprocedural analysis, and symbolic program analysis. Polaris also represents a general infrastructure for analyzing and manipulating Fortran programs, which can provide useful information regarding the program structure and its potential parallelism. The data files used as input to our tool are mainly obtained from the use of Polaris and its infrastructure throughout the program compilation and performance measurement process.

• Parallel programming methodology and tool support
• Program optimization and parallelization environment
• Integrated browsers and anlyzer for program, compilation, and performance data
• Intuitive graphical representation of various types of data
• Data manipulation capability
• File management utilities
• Support for comprehensive comparison and analysis of wide range of reserach data
• Database operation and visualization facilities
• Performance Advisor

The figure bellow shows an overall view of the interaction between Ursa Minor and various data files;

The Ursa Minor tool collects and combines information from various sources. These include static information (data dependence analysis results, interprocedural and IO analysis results, etc.), runtime data (runtime environment, performance results, etc.). The tool allows any type of data to be stored and maintained within itself, expanding its use significantly. Ursa Minor includes facilities for effectively representing these data in graphic viewers for manipulation and analysis. Users can derive new values from raw performance data using the Expression Evaluator. File management utilites helps programmers cope with numerous data files and listings. Also, the Merlin performance advisor has been incorporated into the tool.

Internally, Ursa Minor stores the information in itw own "database". Due to its versatility, it also serves as a tool for comperhensive analysis and comparison in a wride area of reserach. Furthermore, the information can be saved in a format that commercial spreadsheets can read from, allowing a richer set of data manipulation and graphical representations.

The Ursa Minor tool is written in Java with the help of Java Workshop. Thus, any platform on which the Java runtime environment is available can be used to run the tool. It uses the basic Java language with standard APIs, which enhances the portability of the tool. Object orientation in Java allows a relatively easy addition of new types of data that we may want to include in the database. The windowing toolkits and utilities provided a good environment for prototyping user interfaces, which enabled us to focus on the design of the tool functionality.
 

As shown in the above figure, the Ursa minor tool presents the gathered information mainly through two display windows: A table view and a call graph view.  The interface between a user and the tool takes place through menus, mouse, or keyboard.


Click here to view the full size picture.

The above picture shows a typical table view. Here, the table displays information such as timing results from various program runs and the number of invocation of loops. Whenever additional information is available, more columns can be added. Users can perform table manipulation, such as sorting, arranging, and deleting columns and rows. Bar graphs, diplay thresholds, and other utilities add to functionality of the view. Users also have access to the performance advisor and the Expression Evaluator from this view.


click here to view the full size picture.






The above picture presents the call graph view. Each rectangle represents a program unit (subroutine, function, loop). Users can zoom in and out, find units, or have the tool display the portion of source code.

For more information on the functionality, refer to the online User's Manual.

We are currently working on more extensive support for the optimization and parallelization methodology.

The Ursa Minor project is a basis for another project going on at Purdue University. The Ursa Major project (User Responsive System allowing Managed Access to Java Optimization Repository), utilizes the APIs developed for Ursa Minor for the purpose of distributing our results and experience through the Internet. For more information, try the Ursa Major project home page.

Ursa Minor has been installed on the Parallel Programming Hub on Purdue University Network Computing Hub (PUNCH). Any remote user can try out Ursa Minor using this network computing facility. For more information,
try the Parallel Programming Hub.

Related publications can be found on the author's page.
If you have any questions or if you are interested in trying Ursa Minor, please send me an e-mail.