精华公布栏

发信人: carl.bbs@bbs.swjtu.edu.cn (师哥), 信区: cnunix
标  题: 曙光1000 知识讲座（二）
发信站: 锦城驿站 (Sat Dec  6 20:37:36 1997)
转信站: Lilac!ustcnews!ustcnews!sjtunews!swjtu!swjtubbs
出  处: xnjd21.swjtu.edu.cn

Note: 曙光1000在英文环境下运行，所以其资了也以英文的为主。

1. Overview  of Dawning-1000

        The Dawning-1000 Massively Parallel Processing System, which is the
most important key project of the 863 Program, has been successfully developed
by the National Research Center for Intelligent Computing Systems (NCIC) in
April, 1995. Its success indicates that the technologies of the massively
parallel processing (MPP) of China have come up to the advanced world levles,
and its success is considered as a milestone of the high performance computer
community of China.

        The Dawning-1000 broke through some key technologies in the MPP, and
exploits a lot of advanced technologies of 1990s. Its peak speed achieves 2.5
GFLOPS (billion floating point instructions per second), and its practical
computing speed is 1.58 GFLOPS. It has up to 1024 Mbytes of memory. The
Dawning-1000 represents the most advanced high-performance computer in China.
It is a efficient, stable and reliable system which is easy to use. Its
technical features includes:

***      Good extensibility, include scalable connection network, expandable
computing power and I/O power. Both its hardware and software are scalable. It
can scale up to a MPP system containing hundries or even thousands of computing
nodes. It obviously excels the SMP systems and the tranditional vector machines
in this point.

***     Large number of nodes, high computing power, large volume of memory.
It can handle the problems which are hard to deal with by workstations or
traditional supercompter. For examples, a weather forecast program, which
forecasts the coming 48 hours weather condition of the whole area of China,
needs only 3 hours to run on the Dawning-1000, but runs over 60 hours on
ordinary computers such as workstations (the results are of no use then). A
3-dimention iterative deviation kernel program for seismic data processing in
the petroleum prospecting field runs only 11 minutes on the Dawning-1000,
but runs over 10 hours on workstations without results. Compares with vector
supercomputers, the Dawning-1000 has higher performance/cost ratio.

***     The self-designed, stable and reliable wormhole communication chip,
which extensively exploits the combination of asynchrous mode and synchrous
mode, with higher message passing speed.

***     The message passing protocol with flow control, which provided high
efficiency and high reliability for system communications.

***     High quality of the optimizing compiler, which provided efficient
and practical parallel compile and comes up to the advanced world level.

***     the advanced parallel programming environment, parallel debugging
tools, automatic parallelizing tools and the realtime visualization tool of
parallel programs, including PVM, EXPRESS, NX Interface, Autopar, NDB, and
Paravision, etc., make the use of the MPP system much easier.

***     Standard and high quality of the system design and the hardware
production, high quality of the system software, high reliability of the whole
system.

***     Extensively exploiting the international standards, the highly open
system, which makes it much easier for the users to port the environments,
tools, and large application softwares.

The Dawning-1000 comes up to advanced world level of the early 1990s.

        The Dawning-1000 can be widely used in many fields such as seismic
data processing in the petroleum prospecting field, weather forecast,
scientific research, teaching and education, business processing, banking,
stock broking, and so on, it can bring about tremendous economic and social
benefits. Especially, It is the ideal choice for the computer system of
supercomputing centers.

--

  三人同行   必有我师
       倘佯校网    愿交吾友
          有朋自网上来,幸哉、喜哉...
          永远的朋友Carl

※ 来源:．锦城驿站 bbs.swjtu.edu.cn．[FROM: 202.115.65.79]