Refine your search:     
Report No.
 - 
Search Results: Records 1-3 displayed on this page of 3
  • 1

Presentation/Publication Type

Initialising ...

Refine

Journal/Book Title

Initialising ...

Meeting title

Initialising ...

First Author

Initialising ...

Keyword

Initialising ...

Language

Initialising ...

Publication Year

Initialising ...

Held year of conference

Initialising ...

Save select records

Oral presentation

High-performance parallel diagonalization for huge matrices in quantum lattice-fermion problems

Yamada, Susumu; Imamura, Toshiyuki*; Kano, Takuma; Machida, Masahiko

no journal, , 

no abstracts in English

Oral presentation

Transient thermal stress analysis of a spherically curved tubesheet

Hazama, Osamu

no journal, , 

At CCSE, JAEA, a new notion of assembly structure analysis within a Grid computing environment is proposed and a finite element structural analysis software called FIESTA was designed and implemented based on this method. The FIESTA solver is able to "glue" those data made independent of one another using a penalty. Since the introduction of a penalty degrades the condition of the global stiffness matrix, the effect of the parameter on the convergence of the conjugate gradient solver was investigated through a simple sample problem. It was witnessed that as the size of the penalty exceeds the elastic modulus, the rate of increase in the procedure time accelerates. Therefore, there is a trade-off between computational time and accuracy, which must be decided at the time of simulation. Using FIESTA, it is anticipated that a feasible design will be developed for the new-concept tubesheet structure considered under the development of a new FBR.

Oral presentation

One approach of new parallel architecture for real space descritization methods

Kushida, Noriyuki

no journal, , 

In this presentation, a new approach of new computer architecture specialized for real space descritization method was presented. To achieve the speed up of scientific supercomputer, there are roughly two ways; (1) Accelerate processing speed of each unit (2)Increase the number of precessing unit. Currently, we are reaching the bound of both ways, however, I could point out thatone can extremely reduce the amount of communication when the algorithm is specialized to real space descritization method. In line with this context, memory connected parallel architecture was introduced and I could present that principal numerical operation could be operated on such architecture.

3 (Records 1-3 displayed on this page)
  • 1