GPU Support and scalability analysis on the DSL Devito
Exploration geophysics , domain specific language, high performance computing, parallel computing, devito
In the Oil and Gas industry, seismic exploration is a field full of challenges. To achieve better results and reduce costs, techniques are developed and sharpened, often with high complexity and computational demand. To facilitate the development and utilization of those techniques many domain-specific languages(DSL) have been created, seeking to abstract a top-level representation that bridges the many varied scientific fields utilized when applying such techniques. One of these DSLs is Devito and in this research, we take a look at ways of improving it. First by combining it with another tool to provide GPU support in Devito, and secondly by combining it with powerful parallel python libraries seeking to parallelize the execution of code in parallel distributed environments. Devito with GPU capabilities is tested by a wave propagation model and the parallelization is planned to be tested with the full wave-form inversion technique. With such additions to the already powerful Devito optimizations, we plan on reducing execution time and increasing speedup and parallel efficiency.