Solving very large elliptic problems on a supercomputer with Solid State Disk

Journal Article · Wed Aug 01 04:00:00 EDT 1984 · J. Comput. Phys.; (United States)

DOI:https://doi.org/10.1016/0021-9991(84)90011-1· OSTI ID:5380695

Bucher, I Y; Jordan, T L

An experiment is described that demonstrates how secondary solid state storage devices extend effectively the capabilities of supercomputers to solve large computational problems. A linear system of size 45/sup 3/ resulting from the 27-point finite difference discretization of a three-dimensional elliptic operator is solved by a conjugate gradient algorithm on one processor of a Cray X-MP with attached 8M-word Solid State Disk (SSD). The main matrix was stored on the SSD. Although for the 8M-work SSD the problem is I/O bound, it is projected that the 32M-word SSD will support the computation at its full measured speed of 161 MFLOPS. It is concluded that the solution of a 90 x 90 x 90 27-point finite difference problem will require about 0.25 sec per iteration on the Cray X-MP, which will make three-dimensional problems of reasonable size feasible. 2 references, 2 tables.

• Research Organization: Los Alamos National Lab., NM
• OSTI ID: 5380695
• Journal Information: J. Comput. Phys.; (United States), Journal Name: J. Comput. Phys.; (United States) Vol. 55:2; ISSN JCTPA
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200* -- Mathematics & Computers
COMPUTERS
CRAY COMPUTERS
DIFFERENTIAL EQUATIONS
DIGITAL COMPUTERS
EQUATIONS
FINITE DIFFERENCE METHOD
ITERATIVE METHODS
MEMORY DEVICES
NUMERICAL SOLUTION
PARTIAL DIFFERENTIAL EQUATIONS
PERFORMANCE
SUPERCOMPUTERS
USES