- Copyright © 2000 Society of Exploration Geophysicists
The advantages of parallel processors are shown by Figure 1 (taken from Kalantzis, 1994), which illustrates scalar, vector, parallel, and vector/parallel processes for putting out a fire with buckets of water. In scalar mode, a single processor performs the operation (one bucket carrier). In vector mode, the task is performed by a series of simultaneous operations on a vector or array of elements (the bucket brigade). In parallel mode, a number of processors perform similar tasks (several bucket carriers running in parallel). The vector/parallel operation has several lines of bucket brigades operating at once. Obviously, the latter would perform the fastest.
The beauty of Kirchhoff depth migration is that individual traces or shot records can be selectively migrated to depth and summed. The aplanatic surfaces for each trace can be generated by computing seismic wavefronts emanating from the source and summing these wavefronts to those wavefronts generated for each receiver. Total reflection traveltime equals the time from source to reflector plus the time from reflector to receiver; hence, the total reflection time map, or “aplanatic surface,” equals the sum of the wavefront traveltimes mapped from the source with those mapped out from the receiver.
Seismic amplitudes are distributed over the aplanatic surfaces for each trace. One should also account for obliquity factors in the amplitude (Scales, 1995). These depth images are summed together for all traces. Since the same source wavefronts are used in the computation of all aplanatic surfaces in a shot gather, it may be convenient to give each processor a shot gather. The computation of shot migrations can be done on individual processors and then summed together to form a depth-migrated image. This computational process has been …