PTLIB Review of Paragraph

See the PTLIB Review of Parallel Debuggers and Performance Analyzers which includes this evaluation of Paradyn for the review criteria and methodology as well as a comparison with other similar tools.

Performance

Acceptable monitoring overhead
N/A

Acceptable response time
Response time is very good.

Memory/disk requirements
It takes up very little disk space and memory.

Scalable data collection
N/A
Scalable data presentation
Yes, you can zoom in or out using a variety of graphs.

Versatility

Languages/programming models/ communication libaries supported

Runs on currently popular platforms
Yes, runs on most systems
Platform dependencies isolated
No.

Support for heterogeneous environment
Not applicable.

Interacts with current or soon-to-be standards (e.g., PVM, MPI, HPF)
It supports the PICL trace format. So PVM and MPI are supported indirectly.

Uses SDDF
No.

Change/customize/add new views easily
No.

Ease of Use

Documentation
Documentation is good and concise but I would have liked to see a tutorial.

Ease of installation Installation is very easy. You unpack the file you get by executing the file which will unpack it into the dir. Then you go into the dir and make with the makefile of your particular architecture.

Command-line interface
N/A

Window-based interface
Yes.

GUI common look-and-feel (OSF/Motif Style Guide)
Yes.

Privilege-free installation
Yes

Reports information at source code level
No.

Automated instrumentation
N/A

Maturity

Runs without crashing the monitored program
N/A

Reports and recovers from error conditions
Yes.

Support

Capabilities

Support for multiple threads per node
Presents different levels of abstraction, from global to individual threads, procedures, or data structures
Yes, through the different graphs you can see a variety of information at different levels.
Single point control for parallel debugging
N/A

Attach/detach to/from running program
No.

Breakpoints and data watchpoints
N/A

Program state examination
N/A

Program state modification
N/A

Event tracing mechanism
Yes, provided by PICL.

Cache and memory reference tracking/display
No.

Remote data access pattern analysis
No.

Message tracing/display
Yes.

Input/output characterization
No.

Real-time monitoring
No.

Post-mortem analysis
Yes.

Profiling at level of subprocedures and coarse blocks
No.

Utilization display (communications/idle/IO/computation)
Yes.

Performance prediction
No.

Comparisons between different runs
No.

Other

Commercial/research
Research

Cost
Free

Webpage
http://www.netlib.org/paragraph/

Version Date
June 1994

Email
heath@ncsa.uiuc.edu

Summary

Paragraph does a good job at obtaining ease of use with functionality. Paragraph does visual animation of message passing on multi-computer architectures. I found it to be very easy to use and very helpful. While it does not have the flexibility of some of the newer performance analyzers, it has all the tools you need to find the big bottlenecks and problems with your program. The only setback is that you are limited to the given views. But they provide many of these views that should be adequate for most purposes. Basically if you need a quick, easy and portable tool to find the big bottlenecks in your program, I would highly recommend this tool. However if you need to tune your program for the best performance possible you might be better off with another tool.

Ratings (Worse 1 ... 5 Better)


Reviewed by Kevin London, london@cs.utk.edu
August 14, 1996