Referee name/PC Member and affiliation: Geoffrey Fox Syracuse University
_____________________________________________________

Paper Number:20

Paper Title:Irregular Parallel Algorithms in Java

Author(s):   Brian Blount Siddhartha Chatterjee Michael Philippsen 

Please Rate the Paper: (Good 1, Fair 2,  Weak 3)

Quality:      1      

Relevance:   2 

Recommendation (Accept 1, Borderline 2, Reject 3)1


Comments: Not really focussed on irregular issues.
In discussion of speed up address effect of Java's poor node performance
What would happen if computation ran at Fortran speed?

The full paper should try to distinguish effects coming from
Java
Object oriented
Nested Parallelism
Use of threads (as opposed to say MPI like approach)

Paper Number:12

Paper Title:Optimizing Irregular HPF Applications using Halos

Author(s): Anonymous

Please Rate the Paper: (Good 1, Fair 2,  Weak 3)

Quality:     1

Relevance:   1

Recommendation (Accept 1, Borderline 2, Reject 3)1

Comments: Somewhat out of fashion and these extensions are very 
unlikely to move into practice. However lessons from hard work like this
could well be important in future research projects and commercial systems


Paper Number:16

Paper Title:Distributed Object-Oriented Method for Particle Simulation

Author(s): Yudong Sun, Zhengyu Liang, and Cho-Li Wang

Please Rate the Paper: (Good 1, Fair 2,  Weak 3)

Quality:   2  

Relevance: 1  

Recommendation (Accept 1, Borderline 2, Reject 3)2

comment: Java and RMI don't really constitute an interesting computing
model as this is just to way to write code and send messages / invoke remote
functions. Figure 1 is claimed model but I don't find this very deep or likely
to apply to many problems. I suspect that reasonble results largely reflect
Java's lousy node performance. One should normalize to single node
performance of one of the excellent C codes developed for this problem.
In spite of this, paper is clearly written and represents a well thought
through application. There will be good Java based computing models in the future
and they could find these preliminary remarks interesting