The CP-PACS Project aims to develop a massively parallel computer designed to achieve high performance for numerical research of the major problems of computational physics. The project further aims at significant progress in the solution of these problems through the application of the parallel computer upon completion of its development.
The planning of the project was started in the summer of 1991. The proposal,
made to the Ministry of Education, Science and Culture,
was approved in the spring of 1992 as one of projects of the Ministry's
"Program for New Development of Academic Research". The Project
formally started in April of 1992, and is to continue for five years, until
March of 1997.
The project will receive about 22 million US$ spread over the five year period.
The funding comes from a special allocation of the Grant-in-Aid of the
Ministry of Education, Science and Culture supporting innovative fundamental
The Center for Computational Physics was founded in April 1992 at University
of Tsukuba to carry out the Project, as well as to promote
research in compuational physics and parallel computer science. The Center
is an inter-university facility open to researchers in academic institutions
The Project currently has 33 members, of which 15 are computer scientists and 18 are physicists, as listed below.
The Projected is headed by Yoichi Iwasaki. The development of the CP-PACS
computer is led by Kisaburo Nakazawa.
The development of the CP-PACS parallel computer involved a close collaboration of the computer scientists and physicists of the Project. Joint discussions were made on the computational needs of the physicists, on the one hand, and the possible architectures which could meet them within the limits of what would be technically feasable, on the other hand. The basic design of the CP-PACS computer was reached as a result of this collaboration.
For manufacturing of the CP-PACS computer, Hitachi Ltd. was selected in the summer of 1992 by a formal bidding process. The Project members have since worked in close collaboration with Hitachi Ltd. to develop the computer.
The first three years of the Project (1992-1994) were spent in the basic
and detailed designs of the CP-PACS computer and their verification through
simulations. Chip fabrication and assembling of parts started in early
1995. The first stage of the CP-PACS computer consisting of 1024 processing
units with a peak speed of 307Gflops was completed in March 1996. An upgrade
to a 2048 system, which reaches a peak speed of 614Gflops, began in August of
1996, and was completed at the end of September 1996.
The principle goal of the project, in its first phase, was the development of the CP-PACS computer. This required both a thorough assessment of the computational requirements of the physics applications as well as research into the possible architectures. An intensive cross-disciplinary effort by the computer scientists and physicists in the project brought forth a design incorporating a number of novel features to achieve high performance, including pseudo vector processing (PVP-SW) on a RISC processor, and fast communication between the processing units (remote DMA).
With the completion of the CP-PACS computer in 1996, research in computer science now aims to gauge its performance in detail, as a base for further development of parallel computer science.
In computational physics the project aims to use the CP-PACS computer for carrying out research in the following areas:
A major goal of the Project is to significantly advance the numerical study
of lattice QCD in particle physics. QCD (Quantum Chromodynamics) is believed
to be the fundamental theory of strong interactions of elementary particles
such as the proton, neutron and pion. Large-scale numerical simulations will
be pursued with the CP-PACS computer in order to verify the theory and
to extract new physical predictions. Important problems also abound in
condensed matter physics (strongly interacting electron systems and high-temperature
superconductivity, ab initio calculations of material properties etc.)
and in astrophysics (gravitational collapse, large-scale structure of the
Universe etc.). Application of the CP-PACS computer to these areas will
also be pursued.