Numerical Astronomy with Parallel Computers
| Project/Area Number |
62540182
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Astronomy
|
|---|
| Research Institution | University of Tokyo |
|---|
Principal Investigator |
SUGIMOTO Daiichiro Dept. Earth Sci. & Astronomy, Coll. Arts & Sci., Univ. of Tokyo, 教養学部, 教授 (10022592)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
ERIGUCHI Yoshiharu Dept. Earth Sci. & Astronomy, Coll. Arts & Sci., Univ. of Tokyo, 教養学部, 助手 (80175231)
KAWAI Toshio Dept. of Physics, Keio University, 理工学部, 教授 (30146724)
|
|---|
| Project Period (FY) |
1987 – 1988
|
| Project Status |
Completed (Fiscal Year 1988)
|
| Budget Amount *help |
¥1,800,000 (Direct Cost: ¥1,800,000)
Fiscal Year 1988: ¥400,000 (Direct Cost: ¥400,000)
Fiscal Year 1987: ¥1,400,000 (Direct Cost: ¥1,400,000)
|
|---|
| Keywords | Parallel computers / Special purpose computers / Numerical astronomy / Gravitational many body problem / 球状星団 / 並列型コンピュータ |
|---|
| Research Abstract |
In astronomical systems typical is the appearance of density contrast under self-gravitation. In numerical simulation for evolution of such systems there are two major difficulties. Firstly, we have to calculate gravitational interactions between all pairs of particles. Secondly, the density contrast brings about gratly different local time scales. The aims of the present research have been to resolve these difficulties by means of parallel computations and to extend a strategy for future astronomical simulations. Analyzed have been such problems mainly for gravitational many body problems and dynamical evolution of globular clusters. Compared have been different types of algorithms such as direct calculation of gravity requiring N^2 steps and tree algorithms allowing NlogN steps, and different types of vector machines and parallel computers such as PAX and Connection Machine. It has been found that the parallelization could only be performed insufficiently for the sophisticated algorithms while the simple direct algorithm requires too much computational steps.
Thus, I have considered a possibility to construct special purpose computers and found that a simple architecture is possible to resolve the two difficulties mentioned above yet to allow a wide range of applications. It is a hardware realizing Aarseth code of independent time steps in a pipeline. It calculates the motion of one particle to update that reaches its next time step in the first place among others. In its earlier half the coordinates of the other particles are extrapolated from Their own independent times. In its later half the gravitational force acting on the updating particle is calculated. The coordinates of different particles flow through the pipeline performing about 50 floating point calculations. It is also easy to set many pipelines in parallel because the interaction between them appears only in the final stage of gathering gravitational forces contributing from different pipelines.
|
Report
(3 results)
Research Products
(6 results)
