PTLIB Review of DAQV

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

Performance

Acceptable monitoring overhead
Yes.

Intrusion compensation
No.

Acceptable response time
Yes, the controls responded quickly.

Memory/disk requirements
Memory requirements are small enough to run the display on an xterm although it did run a bit slower. Disk requirements is almost 6 megs including the executables and documentation.

Scalable data collection
N/A

Scalable data presentation
N/A

Versatility

Languages/programming models/ communication libaries supported
HPF

Runs on currently popular platforms
SGI workstations and Power Challenge/Onyx Solaris workstations.

Platform dependencies isolated
Yes.

Support for heterogeneous environment
Yes.

Interacts with current or soon-to-be standards (e.g., PVM, MPI, HPF)
Interacts with HPF.

Uses SDDF
No.

Change/customize/add new views easily
N/A

Ease of Use

Documentation
Not, much documentation on it. You can get an overview from their web page. There are a few readme files but I found the documentation to be in-adequate.

Ease of installation
Installation is not bad. You must have the pghpf compiler and other that that you are set as they send the tcl, tk libraries with it (which is nice).

Command-line interface
Sample command-line control and data clients are included in the distribution.

Window-based interface
Yes.

GUI common look-and-feel (OSF/Motif Style Guide)
Yes, has a good gui feel and look to it.

Privilege-free installation
Yes, there is no need for special privileges to install this package.

Reports information at source code level
Yes, it can report the code to the source level.

Automated instrumentation
Yes, it can be automated or it can be done manually.

Compile without special linking
No.

Maturity

Runs without crashing the monitored program
Yes, applications don't crash due to DAQV

Reports and recovers from error conditions
Yes, it seems to recover from errors well.

Capabilities

Support for multiple threads per node
No.

Presents different levels of abstraction, from global to individual threads, procedures, or data structures
No.

Single point control for parallel debugging
Yes.

Attach/detach to/from running program
No.

Breakpoints and data watchpoints
Yes, you can stop at certain lines of the program.

Program state examination
Yes.

Program state modification
No, however framework has being extended to support computational steering support in which user subroutines can be invoked via the external control client and also change values of distributed data structures through a similar mechanism.

Event tracing mechanism
No.

Cache and memory reference tracking/display
No.

Remote data access pattern analysis
N/A

Message tracing/display
N/A

Input/output characterization
N/A

Real-time monitoring
Yes.

Post-mortem analysis
No.

Profiling at level of subprocedures and coarse blocks
N/A

Utilization display (communications/idle/IO/computation)
N/A

Performance prediction
No.

Comparisons between different runs
No.

Other

Commercial/research
Research

Cost
Free

Webpage
http://www.cs.uoregon.edu/~hacks/research/daqv/

Version Date
July 1996

Version
1.0

Libraries Required
pghpf Version 2.0 or 2.1 tcl7.4 and tk4.0 (Included for SGI and Solaris)

Summary

DAQV is a tool for distributed array query and visualization in HPF programs. This tool has a lot of potential, but it is not a polished tool. This did not seem to be the main focus of this project , but more aimed at making a tool that can be built upon. It achieves this very well. DAQV's one feature right now is the ability to extract the data values of distributed arrays for analysis or display. The display they provide is bare bones, but DAQV provides ways to to develop different displays. The documentation I thought was not aimed as much towards the programmer but more the parallel tool developers. DAQV does a lot for the tool developer because it supports client/server protocols for direct interaction with executioner HPF code. Overall this tool has lots of potential if it gets built upon.

Ratings (Worse 1 ... 5 Better)

Click here to view a screen shot
Reviewed by Kevin London, london@cs.utk.edu
August 26, 1996