eugene@riacs.edu (Eugene Miya) (12/15/88)
Every year for the past 4 years I have polled researchers for
significant papers in parallel and distributed computing.
This will be the last time I do this. I have decided to discontinue
active support of my parallel processing bibliography.
I have decided to do this for several reasons: 1) I have changed
organizations from a research group into a development group so I
have (and my management has) a different focus, 2) the effort has NOT
proven to be worth the benefit. Bibliographies are typically things
you assign grad students right? I have found some research projects
to be very cooperative (I salute the Ultracomputer people as example),
but I have found several other projects less than helpful (I will
not mention names, but think they either do "shady" work or hold their
cards close for commercial reasons), but I don't blame them.
Economics is one thing which separates computer science from the other
real sciences (graphics possibly be the worst offender).
Anyways, I don't have time to scan the stacks anymore, nor have secretarial
support for key entry. The IEEE has suggested taking over, this might
be a possibly. The problem is they might not do better in the cooperation
area. {Others have suggested soliciting fund for secretarial help,
I don't have time to scan either.}
I began the bibliography as part of NASA funded research for a survey
of parallel processing architectures. It became very clear to me
that the Federal Government doesn't build hardware. Also clear
was that numerous Surveys (excellent works) were getting written, but
no one was reading them. Every Agency, company, etc. with a vested
interest was commissioning surveys, with no one to read them. I decided
to go machine readable (you would not believe some of the inaccuracies
which your colleagues had published in some earlier works, and I
can't vouch I've done much better except that I can correct and
redistribute). The thing grew and grew and will continue to grow.
Will parallel processing become all of computing [full relevance]? Maybe.
There is too much duplication of effort, too much horn tooting,
and I know everyone wants tenure, etc. There is no coordination
in this country and too much negative feeling between "competing groups."
There is no sense of cooperation or compromise between developers and
users. I think there is a lack of foresight. I never felt that I was
able to get enough support from the community with people playing their
information so close.
You can say that is the nature of the world (jumping hoops), but I think
I see more. I see too many people not willing to invest into the long term.
I see quite a few people who don't know what they are talking about,
hell, I frequently have no concept of what I am doing sometimes.
I see a holy grail of automatic parallelism following the grail of
automatic programming of the 1960s.
So anyways, I want to update the suggested readings list one last
time for the benefit of those who can use it. I plan to freeze it
in two weeks (unless I could find a good home which will maintain
a master copy, maybe the IEEE). The ground rules for suggested reading:
what would you have a 1st year grad student read to get started?
The top ten papers in frequency get a "required" reading keyword,
the next 100 get "recommended" reading stamp.
The past has seen some comment that 1) the list is all old architectures:
True, people are generally conservative and prefer to point to real
machines not paper designs. Paper designs take heavy criticism (data
flow as an example). 2) new machines get inadequate coverage (possibly
too many of them). 3) heavy emphasis on hardware without enough
software: justified. If you think there are papers students should read,
which are not on this list, tell me which. Don't just complain.
If others agreed, then enough will get a required keyword.
So what would you have a 1st year grad student read?
I have my personal list, what would you read?
The past list is given below. I don't like this list myself, but I
only count as one vote (which at least gets a recommendation). This
list should not be constructed as "The list," it's "a list."
All copywritten annotations have been scrubbed. Please cite where you got
it if you use it (keeps me employed). When I see one more request
for citations on some parallel processing topic, I plan to 'u' comp.parallel.
I have to get on to REAL research, not maintain a reading list.
Yet another gross generalization from the Rock of Ages Home for Retired
Hackers
--eugene miya, resident cynic
Send suggested readings lists to:
eugene@orville.nas.nasa.gov or {favorite UUCP}!ames!eugene
List: cut here
%A K. E. Batcher
%T STARAN Parallel Processor System Hardware
%J Proceedings AFIPS National Computer Conference
%D 1974
%P 405-410
%K grequired
%X This paper is reproduced in Kuhn and Padua's (1981, IEEE)
survey "Tutorial on Parallel Processing."
%A David J. DeWitt
%A Raphael Finkel
%A Marvin Solomon
%T The CRYSTAL Multicomputer: Design and Implementation Experience
%I University of Wisconsin-Madison
%R Computer Sciences Technical Report #553
%D September 1984
%K grequired
%X A good current overview of the Crystal project. The first part reads like
the C.mmp retrospective by Wulf [1980] et al. They suffered from the same
problems as CMU: small address space, reliability, and they also pushed
the software the software forward to the next stage of problems.
%A Robert H. Kuhn
%A David A. Padua, eds.
%T Tutorial on Parallel Processing
%I IEEE
%D August 1981
%K grequired
%X This is a collection of noted papers on the subject, collected for
the tutorial given at the 10th conference (1981) on Parallel Processing.
It eases the search problem for many of the obscure papers.
Some of these papers might not be considered academic, others are
applications oriented. Data flow is given short coverage. Still, a
quick source for someone getting into the field.
Where ever possible, paper in this bibliography are noted as being in this
text.
%A G. J. Lipovski
%A A. Tripathi
%T A reconfigurable varistructure array processor
%J Proc. 1977 Int. Conf. on Parallel Processing
%D August 1977
%P 165-174
%K grequired, U Texas, TRAC
%A Richard M. Russell
%T The Cray-1 Computer System
%J Communications of the ACM
%V 21
%N 1
%P 63-72
%D January 1978
%K grequired,
existing classic architecture,
maeder biblio: parallel hardware and devices, implementation,
ginsberg biblio:
%X The original paper describing the Cray-1.
This paper is reproduced in Kuhn and Padua's (1981, IEEE)
survey "Tutorial on Parallel Processing."
Also reproduced in "Computer Structures: Principles and Examples" by
Daniel P. Siewiorek, C. Gordon Bell, and Allen Newell, McGraw-Hill,
1982, pp. 743-752.
Reproduced in Dharma P. Agrawal's (ed.) "Advanced Computer Architecture,"
IEEE, 1986, pp.15-24.
%A Richard J. Swan
%A S. H. Fuller
%A Daniel P. Siewiorek
%T Cm* \(em A Modular, Multi-Microprocessor
%J Proceedings AFIPS National Computer Conference
%I AFIPS Press
%V 46
%D 1977
%P 637-644
%K CMU, grequired
%X This paper is reproduced in Kuhn and Padua's (1981, IEEE)
survey "Tutorial on Parallel Processing."
%A Philip C. Treleaven
%A David R. Brownbridge
%A Richard P. Hopkins
%T Data-Driven and Demand-Driven Computer Architecture
%J Computing Surveys
%V 14
%N 1
%D March 1982
%P 93-143
%K grequired,
CR Categories and Subject Descriptors: C.0 [Computer System Organization]:
General - hardware/software interfaces; system architectures;
C.1.2 [Processor Architecture]:
Multiple Data Stream Architectures (Multiprocessors);
C.1.3 [Processor Architecture]: Other Architecture Styles
- data flow architectures; high level language architectures;
D.3.2 [Programming Languages]: Language Classifications - data-flow
languages; macro and assembly languages; very high-level languages
General Terms: Design
Additional Key Words and Phrases: Demand = driven architecture,
data = driven architecture
%X * The aim of this paper is to identify the concepts and relationships
that exist both within and between the two areas of research of
data-driven and demand-driven architectures.
Reproduced in "Selected Reprints
%A W. A. Wulf
%A C. G. Bell
%T C.mmp \(em A multi-mini processor
%J Proc. Fall Joint Computer Conference
%V 41, part II
%I AFIPS Press
%C Montvale, New Jersey
%D December 1972
%P 765-777
%K multiprocessor architecture and operating systems
grequired,
parallel processing,