high performance Java mailings
David G Hough at validgh
validgh
Tue Jul 21 22:02:09 PDT 1998
Date: Mon, 20 Jul 1998 13:46:56 -0400
From: "SF.Hummel" <hummelawatson.ibm.com>
Subject: CFP
To: java-for-cseanpac.syr.edu
#######################################################################
# #
# Scientific Programming #
# #
# Call for Papers #
# A Special Issue Focusing on #
# High Performance Java Compilation and Runtime Issues #
# #
#######################################################################
*** Introduction to the topic/motivation***
Due to the limited life-span of today's supercomputers, achieving
performance portability of parallel codes is of critical importance.
Although Java was not specifically designed for the computationally
intensive numeric applications that are the typical fodder of
supercomputers, its widespread popularity and portability make it an
interesting candidate for high-performance computing; however, many
open research questions remain regarding the viability of high
performance Java-enabled environments. These encompass topics ranging
from compiler/runtime optimizations to the language design to
interfacing with legacy libraries.
Papers are solicited for a Special Issue of Scientific Programming,
tentatively scheduled to be published in July, 1999.
The purpose of this special issue is to focus on the fast-paced
developments, experiences and research supporting high-performance
execution of Java codes for numerical applications.
Topics of interest include (but are not limited to):
*** your list of topics
o JITs/Dynamic compilation
o Native code generation
o Analysis of numeric applications
o Compiler optimizations
o Garbarge collection
o Language Specification (floating-point, array, object and
thread models)
o JVM design and implementation
o Scalability issues
o Highly parallel implementations
o Distributed/web-based implementaions
o Interfacing with legacy libraries
Schedule:
--------------------------------------------------
| Submission deadline: July 31, 1998
| Notification of Acceptance: December 31, 1998
| Target month of Special Issue: July 1999
--------------------------------------------------
Authors should follow the Information for Authors at the end of each
issue of Scientific Programming when preparing their manuscripts for
submission. An electronic copy of the manuscript, in Postscript format
and viewable with ghostview, should be submitted via electronic mail
to the guest editor by July 31, 1998. Submissions
should include authors names, affiliations, addresses, fax and phone
numbers, email addresses, on the cover page. Only original manuscripts
will be considered. All papers will undergo a peer review
process. Authors
will be notified of the final publication decision by December
31, 1998.
#######################################################################
# Guest Editor, Scientific Programming #
# #
# Susan Flynn-Hummel #
# IBM T. J. Watson Research #
# 30 Saw Mill River Road #
# Hawthorne, NY 10532 #
# Tel: (914) 784-7942, Fax: (914) 784-6242 #
# email: hummelawatson.ibm.com #
# #
#######################################################################
------------- End Forwarded Message -------------
------------- Begin Forwarded Message -------------
Date: Tue, 21 Jul 1998 13:45:14 -0500 (CDT)
From: Corky Cartwright <corkacs.rice.edu>
To: java-for-cseanpac.syr.edu
Subject: A Much Higher Performance Java Implementation for Solaris 2.6
Solaris 2.6 Users:
Sun has just released a new implementation of Java 1.2 designed
specifically for Solaris machines (Sparc and x86) under its
Early Access program at the Java Developers Connection. It is
available at the URL
http://developer.javasoft.com/developer/earlyAccess/jdk1.2/index.html
You have to register with the Java Developers Connection to download
the system but the registration process is trivial.
The implementation features a completely new generational copying
garbage collector, low overhead locking protocols, and a much more
sophisticated JIT compiler (it even does some common subexpression
elimination). During the past academic year, I spent my sabbatical at
Sun Microsystems Laboratories East working with Java Topics research
group that produced this implementation. The experimental Java
implementation developed by this research group was so better than
existing Sun products that the Solaris Sofware division of Sun
Microsystems decided to make the research system into a product.
I am interested in hearing how well this implementation performs on
scientific calculations. It is still a long way from optimized
Fortran, but it runs 12 times faster than JDK 1.1.5 on my favorite
symbolic benchmark (a boolean simplifier).
Check it out.
Corky Cartwright
Department of Computer Science
Rice University
More information about the Numeric-interest
mailing list