RACES'12: Relaxing Synchronization for Multicore and Manycore Scalability, SPLASH'12 Workshop

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

RACES'12: Relaxing Synchronization for Multicore and Manycore Scalability, SPLASH'12 Workshop

Stefan Marr-5
==============================================================================
                            Call for Participation


                             R A C E S    2 0 1 2

       Relaxing Synchronization for Multicore and Manycore Scalability

              Workshop Co-located with SPLASH in Tucson, Arizona
                              Sunday, October 21

                    Submission deadline: Monday, August 6 
                         http://soft.vub.ac.be/races/

==============================================================================



Massively-parallel systems are coming: core counts keep rising - whether
conventional cores as in multicore and manycore systems, or specialized cores
as in GPUs. Conventional wisdom has been to utilize this parallelism by
reducing synchronization to the minimum required to preserve determinism - in
particular, by eliminating data races. However, Amdahl's law implies that on
highly-parallel systems even a small amount of synchronization that introduces
serialization will limit scaling. Thus, we are forced to confront the
trade-off between synchronization and the ability of an implementation to
scale performance with the number of processors: synchronization inherently
limits parallelism. This workshop focuses on harnessing parallelism by
limiting synchronization, even to the point where programs will compute
inconsistent or approximate rather than exact answers.


Organizers:

Andrew P. Black, Portland State University
Theo D'Hondt, Vrije Universiteit Brussel
Doug Kimelman, IBM Thomas J. Watson Research Center
Martin Rinard, MIT CSAIL
David Ungar, IBM Thomas J. Watson Research Center


Theme and Topics
----------------

A new school of thought is arising: one that accepts and even embraces
nondeterminism (including data races), and in return is able to dramatically
reduce synchronization, or even eliminate it completely. However, this
approach requires that we leave the realm of the certain and enter the realm
of the merely probable. How can we cast aside the security of correctness, the
logic of a proof, and adopt a new way of thinking, where answers are good
enough but not certain, and where many processors work together in parallel
without quite knowing the states that the others are in? We may need some
amount of synchronization, but how much? Or better yet, how little? What
mental tools and linguistic devices can we give programmers to help them adapt
to this challenge? This workshop focuses on these questions and related ones:
harnessing parallelism by limiting synchronization, even to the point where
programs will compute inconsistent or approximate rather than exact answers.


This workshop aims to bring together researchers who, in the quest for
scalability, have been exploring the limits of how much synchronization can be
avoided. We invite submissions on any topic related to the theme of the
workshop, pro or con. We want to hear from those who have experimented with
formalisms, algorithms, data structures, programming languages, and mental
models that push the limits. In addition, we hope to hear from a few voices
with wilder ideas: those who may not have reduced their notions to practice
yet, but who have thoughts that can inspire us as we head towards this
yet-uncertain future. For example, biology may yield fruitful insights. The
ideal presentation for this workshop will focus on a grand idea, but will be
backed by some experimental result.


Submission
----------

Authors are invited to submit short position papers, technical papers, or
experience reports. Submissions may range from a single paragraph to as long
as desired, but the committee can only commit to reading one full page.
Nonetheless, we expect that in many cases reviewers will read farther than
that. Submissions should be formatted according to the ACM SIG Proceedings
style at http://www.acm.org/sigs/publications/proceedings-templates and should
be submitted via EasyChair at
http://www.easychair.org/conferences/?conf=races2012 in PDF format.


PLEASE NOTE: All submissions (except for those retracted by their authors)
will be posted on the workshop website, along with reviews, which will be
signed by the reviewers, and a rating assigned by the program committee.
Further, the submissions to be presented at the workshop will be selected by a
vote of all registered attendees. As well, submissions to be published in an
official proceedings will be selected by the program committee. Please see the
sections below concerning the rationale and details for this process.


Program Committee
-----------------

Andrew P. Black, Portland State University
Yvonne Coady, University of Victoria
Tom Van Cutsem, Vrije Universiteit Brussel
Theo D'Hondt, Vrije Universiteit Brussel
Phil Howard, Portland State University
Doug Kimelman, IBM Thomas J. Watson Research Center
Eddie Kohler, Harvard SEAS
Jim Larus, Microsoft Research
Stefan Marr, Vrije Universiteit Brussel
Tim Mattson, Intel
Paul McKenney, IBM
Hannes Payer, University of Salzburg
Dan Prenner, IBM
Lakshmi Renganarayana, IBM
David Ungar, IBM Thomas J. Watson Research Center

- TBC -


Important Dates
---------------

August     6     Submission deadline.
August    29     Reviews sent to authors.
September  3     Last date for retraction by authors.
September  4     Papers, reviews, ratings posted on web site. Voting opens.
September 11     Voting closes.
September 14     Notification of papers accepted for presentation and/or publication.
August    21     SPLASH early registration deadline.
October   21     Workshop.
mid-November     Camera-ready copy due for papers selected for proceedings.


Goals and Outcomes
------------------

We will consider the workshop a success if attendees come away with new
insights into fundamental principles, and new ideas for algorithms, data
structures, programming languages, and mental models, leading to improving
scaling by limiting synchronization, even to the point where programs will
compute inconsistent or approximate rather than exact answers. The goal of
this workshop is both to influence current programming practice and to
initiate the coalescence of a new research community giving rise to a new
subfield within the general area of concurrent and parallel programming.
Results generated by the workshop will be made persistent via the workshop
website and possibly via the ACM Digital Library.



  The RACES 2012 Review Process and Workshop Presentation Selection Process
  =========================================================================
  
                                 David Ungar
                     IBM Thomas J. Watson Research Center
                     PC Chair for Workshop  Presentations

Technology has changed the economic tradeoffs that once shaped the reviewing
process. It has become cheap and easy to share submissions, reviews and the
preferences of the attendees. What remains scarce is the number of hours in a
day, and as a consequence the time we have in our workshop in which to learn
and share with each other. I believe that this change in the balance of
factors affords us the opportunity to significantly improve the review and
selection processes.


Sadly, all too often, those who spend their precious time attending a workshop
are not served as well they could be with respect to enlightenment, thought
provoking discussions, and being challenged by new ideas. The fault lies not
in the people who generously donate their time to serve on program committees
and do external reviews. Rather, the fault lies in the process itself. The
very notion of acceptance by committee forces us to boil a rich stew of
reactions, insights, and opinions, down to a single carrot. As a result, it is
common for PC members to come away from a meeting feeling that either some
fraud will be perpetrated on the audience by a fundamentally flawed paper, or,
more often, feeling that a sin of omission will be committed on the audience
by the suppression of a significant but controversial new idea. Sometimes
instead of a carrot we get a lump of gristle.

There are other, lesser, flaws in this process. Although reviewer anonymity
protects negative reviewers from resentment and reprisal, all too often it
prevents an open debate that would promote mutual understanding. Further, in
some cases anonymity allows a reviewer to cast aspersions on authors without
being accountable. Finally, we fail to take maximal advantage of the time and
effort spent in creating insightful reviews when we withhold them from the
audience. Attendees and readers could benefit from expert reactions as they
try to glean the wisdom embedded in the authors' papers.

In this workshop, we have an opportunity to try a different process, one that
we hope will serve all parties better: All reviews will be signed, all
submissions and reviews will be posted on the web (unless an author chooses to
retract a submission), and the attendees will be the ones selecting which
papers will be presented.

Here are the details:
---------------------

At least three committee members will review each submission, and each review
will be signed. Once all the reviews for a submission are in, they will be
sent to the author, who can decide to retract the paper if so desired. Then,
all submissions (except any that are retracted) will be posted on the workshop
website, along with all reviews and a net score determined for each submission
by the program committee.

At this point, prior to the workshop, all registered attendees will be invited
to read the submissions and the reviews, and vote on which of the papers they
want to see presented. Of course, an attendee who so wishes will be free to
merely vote according to the recommendation of the PC, or to not vote and to
accept the wisdom of the rest of the attendees. But the important point
remains: it will be those who will be spending the time in the room who get to
decide how that time is spent. Please note that a submission being posted on
the workshop website and/or presented at the workshop are not intended to
constitute prior publication for purposes of publishing in other workshops,
major conferences, or journals.

This process is a grand experiment, designed to exploit the technologies we
Computer Scientists have created, in order to better serve the advancement of
Computer Science. We hope that its potential excites you as much as it excites
us!



         The RACES 2012 Published Proceedings Paper Selection Process
         ============================================================

                                 Theo D'Hondt
                          Vrije Universiteit Brussel
                        PC Chair for Proceedings Papers

We understand that many submitters may want to publish their paper in an
official proceedings in addition to having it posted on the workshop website.
In order to satisfy that desire, we will publish a proceedings via the ACM
Digital Library. To satisfy ACM DL selectivity requirements, a separate and
more conventional process will be employed for selecting papers to be included
in the published proceedings: Even though all submissions will be posted on
the workshop website (unless retracted by the author), the program committee
will select a smaller number of papers to be included in the published
proceedings based on the signed and posted reviews. Authors of the selected
papers will be asked to submit revised and extended papers mid-November,
taking into account the reviews and the publisher's guidelines. Page limits
for the revised and extended papers to be included in the published
proceedings are anticipated to be 10 pages for research papers, and 5 pages
for position papers. Please note that inclusion in the ACM Digital Library
published proceedings may well be considered to be a prior publication for
purposes of publication in other workshops, major conferences, or journals.
For that reason, authors may choose to decline to have their submission
included in the published proceedings, even if it was presented at the
workshop.



              For questions please contact: [hidden email]
              For updates, follow us on Twitter: @races_workshop


==============================================================================