ALOMA, an auto-parallelization tool for seismic processing
For an efficient execution of seismic processing algorithms on todays hardware architectures parallelization is mandatory. But the corresponding programming effort is high. General aspects like job scheduling and monitoring, data splitting and aggregation or resources management, have to be taken into account, which are inherent to a parallel execution. We present the programming framework ALOMA, a highly efficient autoparallelization tool for a wide range of seismic processing methods. It allows to combine processing modules to complex workflows, which are executed concurrently parallelized over nodes and cores. Control and data dependencies are generated automatically. ALOMA hides the complexity of a generic parallelization framework behind a thin domain-specific interface. An easy to use plug-in architecture allows to add user-defined algorithmic modules. Existing libraries and binaries can easily be integrated without any change.