	SSP 1994 project summary:

[EPCC home] [SSP home] [2001 projects] [2000 projects] [1999 projects] [1998 projects] [1997 projects] [1996 projects] [1995 projects] [1994 projects] [1993 projects]

Exploiting Trace Based Parallelism in Geophysical Calculations

Euphrates as it currently stands is designed to perform image-processing style operations on 3D volumes of data. Hence the name of its commercial re-implementation 3DVIP. However a 3D volume of seismic data represents the final product of a long line of seismic data processing operations.

Seismic, and indeed almost all geophysical, data is recorded as traces: signals recorded in the time domain from individual receivers, such as geophones, hydrophones, seismometers, electrodes, search coils, etc. In order to improve signal to noise ratios and to ensure adequate spatial sampling it is typical to record as many traces as feasible.

Subsequent processing of these traces to produce a final data volume is trace based. In seismic data processing some operations, such as deconvolution, are applied independently to each trace. Others involve accessing a greater or lesser number of other traces e.g. amplitude versus offset (AVO) analysis, which involves the joint analysis of a small collection of traces, and migration, which has proved very difficult to parallelize because, in theory, every input trace contributes to the value of every output trace. In addition there are a number of operations which essentially consist of summing, sorting, warping and re-ordering traces.

The objective of this project is to investigate the parallelisation of trace based operations in the Euphrates framework.

Peilin Jia worked on this project.

Compressed PostScript of Peilin's final report is available here (238245 bytes) .

Webpage maintained by mario@epcc.ed.ac.uk