
Full text loading...
In this paper, we discuss the computational costs of prestack reverse time migration, specially in a parallel environment, where the amount of storage required is high, due to the number of shots being processed simultaneously. This storage space is necessary to record the sections generated on modeling phase, which will be used later on the imaging condition. We show that it is possible to reduce the free space requirement, by saving only few sections spaced in time. The discarded sections can then be later restored with the modeling being held backwards in time. Another approach presented is the spatial decomposition of the section, resulting on a smaller amount of data processed at each computing node.The tests were done with 2-D and 3-D models, using the parallel interface MPI (Message Passing Interface). Results indicate that it is possible to drastically reduce the storage required, keeping the quality of the migrated section, however, with some increase in processing time.