Reducing Data Transfer in Parallel Processing of SQL Window Functions

Parallel processing is an effective computation in which many calculations are carried out simultaneously. In this paper, effects of shared-memory parallel processing on Large Eddy Simulations (LES) in ANSYS Fluent are presented. Fluent provides parallel processing to improve the speed of running programs. LES is one of the most popular viscosity models for turbulence used in computational fluid dynamics (CFD). Three kinds of LES with different sub-grid turbulence models were evaluated: Smagorinsky-Lilly Model (Lilly model), Wall-Adapting Local Eddy-viscosity Model (WALE model) and Wall Modeled Large Eddy Simulation (WMLES model). The running speed of the different models simulating a square duct on a single computer are compared. The relationship between wall-clock time and number of processors reveals the performances of different LES models. The part of the time that is not parallelizable such as file IO and data transfer is also considered based on Amdahl’s law.

Download Full-text

Proposal of Parallel Processing Area Extraction and Data Transfer Number Reduction for Automatic GPU Offloading of IoT Applications

Lecture Notes in Computer Science - Smart Computing and Communication ◽

10.1007/978-3-030-05755-8_5 ◽

2018 ◽

pp. 39-54 ◽

Cited By ~ 1

Author(s):

Yoji Yamato ◽

Hirofumi Noguchi ◽

Misao Kataoka ◽

Takuma Isoda ◽

Tatsuya Demizu

Keyword(s):

Parallel Processing ◽

Data Transfer ◽

Transfer Number ◽

Iot Applications

Download Full-text

Development of PC Cluster Parallel Processing Scheme of 3-D Rigid-Plastic Finite Element Method Using Diagonal Matrix

Key Engineering Materials ◽

10.4028/www.scientific.net/kem.340-341.371 ◽

2007 ◽

Vol 340-341 ◽

pp. 371-376 ◽

Cited By ~ 1

Author(s):

Kenichiro Mori ◽

Y. Kanno

Keyword(s):

Finite Element Method ◽

Finite Element ◽

Parallel Processing ◽

Distributed Memory ◽

Data Transfer ◽

Diagonal Matrix ◽

Rigid Plastic ◽

Pc Cluster ◽

Memory Type ◽

Element Method

The 3-D rigid-plastic finite element method using a diagonal matrix was applied to parallel processing using a distributed memory type PC cluster. The cluster composed of cheap PCs becomes common as a low-cost system in the parallel processing. Since the computers in the distributed memory type PC cluster have individual memory units, the transfer of date among computers during the computation is required, and thus the time for the data transfer is taken into consideration. The renewal of data in each computation is limited because of the time of data transfer unlike the shared memory type workstation. This brings about the delay of data renewal. A data transfer scheme was investigated to optimize the total computational time in the parallel processing. The effect of the delay of date renewal on the convergence of the solution was examined in the simulation of upsetting of rectangular block with an inclined tool by means of a cluster composed of 4 PCs and 100MBit/s Ethernet.

Download Full-text

Holistic Shuffler for the Parallel Processing of SQL Window Functions

Distributed Applications and Interoperable Systems - Lecture Notes in Computer Science ◽

10.1007/978-3-319-39577-7_6 ◽

2016 ◽

pp. 75-81 ◽

Cited By ~ 2

Author(s):

Fábio Coelho ◽

José Pereira ◽

Ricardo Vilaça ◽

Rui Oliveira

Keyword(s):

Parallel Processing ◽

Window Functions

Download Full-text

Study of parallel processing area extraction and data transfer number reduction for automatic GPU offloading of IoT applications

Journal of Intelligent Information Systems ◽

10.1007/s10844-019-00575-8 ◽

2019 ◽

Vol 54 (3) ◽

pp. 567-584

Author(s):

Yoji Yamato

Keyword(s):

Parallel Processing ◽

Data Transfer ◽

Transfer Number ◽

Iot Applications

Download Full-text

SPECTRA: A program for processing electron images of crystals

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100121065 ◽

1992 ◽

Vol 50 (1) ◽

pp. 132-133

Author(s):

M.F. Schmid ◽

R. Dargahi ◽

M. W. Tam

Keyword(s):

Lattice Distortion ◽

Data Transfer ◽

Reciprocal Lattice ◽

Data Entry ◽

Image Data ◽

Distortion Correction ◽

Specimen Preparation ◽

Electron Image ◽

Computer Controlled ◽

Program Modules

Electron crystallography is an emerging field for structure determination as evidenced by a number of membrane proteins that have been solved to near-atomic resolution. Advances in specimen preparation and in data acquisition with a 400kV microscope by computer controlled spot scanning mean that our ability to record electron image data will outstrip our capacity to analyze it. The computed fourier transform of these images must be processed in order to provide a direct measurement of amplitudes and phases needed for 3-D reconstruction.In anticipation of this processing bottleneck, we have written a program that incorporates a menu-and mouse-driven procedure for auto-indexing and refining the reciprocal lattice parameters in the computed transform from an image of a crystal. It is linked to subsequent steps of image processing by a system of data bases and spawned child processes; data transfer between different program modules no longer requires manual data entry. The progress of the reciprocal lattice refinement is monitored visually and quantitatively. If desired, the processing is carried through the lattice distortion correction (unbending) steps automatically.

Download Full-text