Parallel implementation

The methodology described above is not feasible on current archetecture. The computational requirements are onerous, but potentially managable. However, the memory requirements are not. A full 5-dimensional cube, that we are creating when applying $\bf L$, can easily achieve tens of gigabytes. This size of data makes it almost impossible to practically implement any algorithm for 3-D prestack seismic data-processing on a single machine.

Clapp (2004) introduces an efficient python library for handling parallel jobs. The library makes it easy for the user to take an already existing serial code and transform it into a parallel code. The library handles distribution, collection, and node monitoring, commonly onerous tasks in parallel processing.

The main prerequisite to using the python library is to build an efficient serial code, and to describe how the parallel job should be distributed on a cluster. For our problem we chose to split along the h_x axis. We create a series tasks, each assigned to produce a single (${\rm time}, {\rm cmp}_x, {\rm cmp}_y$) volume. Each task is passed a range of h_x's defined by equation 8. The resulting image volumes are then recombined to form the 4-D output space.