Reverse Time Migration (RTM) is a state-of-the-art algorithm used in seismic depth imaging in complex geological environments for the oil and gas exploration industry. It calculates high-resolution images by solving the three-dimensional acoustic wave equation using seismic datasets recorded at various receiver locations. Using a finite-difference time-domain (FDTD) scheme, the time integration follows an adjoint-state formulation with two successive phases, i.e., forward modeling and backward integration. Each subsurface image is then generated during the imaging condition that combines a forward propagated source wavefield with a backward propagated receiver wavefield. RTM’s computational phases are predominantly composed of stencil computational kernels for the FDTD, applying the absorbing boundary conditions, and I/O operations needed for the imaging condition. In fact, RTM can be considered as an out-of-core algorithm, which requires offloading to disk snapshots of the domain solution at specific time intervals during the forward modeling phase. During the backward time integration, these snapshots are read back and synchronized at the corresponding times. As far as optimizing the stencil computation, spatial blocking represents the widely-adopted vendor-agnostic technique for increasing data reuse in the high-level of the memory subsystem. In this paper, we integrate in RTM the asynchronous Multicore Wavefront Diamond (MWD) tiling approach that permits to further increase data reuse by leveraging spatial with Temporal Blocking (TB) during the stencil computations. This integration engenders new challenges with a snowball effect on the legacy synchronous RTM workflow as it requires to deeply rethink of how the absorbing boundary conditions, the I/O operations, and the imaging condition operate. These disruptive changes in cascade are necessary to maintain the performance superiority of asynchronous stencil execution throughout the time integration, while ensuring the quality of the subsurface image does not deteriorate. We assess the overall performance of the new MWD-based RTM and compare against traditional SB-based RTM on various shared-memory systems using the SEG Salt3D model. The MWD-based RTM is able to achieve up to 60% performance speedup compared to SB-based RTM. To our knowledge, this paper highlights for the first time the applicability of asynchronous RTM executions, which results in a higher simulation throughput and may eventually create new research opportunities in improving the hydrocarbon extraction for the petroleum industry.
|Original language||English (US)|
|Title of host publication||35th IEEE International Parallel & Distributed Processing Symposium|
|State||Published - Nov 1 2020|
Bibliographical noteKAUST Repository Item: Exported on 2020-11-02
Acknowledgements: The authors would like to thank Cray Inc. and Intel in the context of the Cray Center of Excellence and Intel Parallel Computing Center awarded to the Extreme Computing Research Center at KAUST. For computer time, this research used Shaheen-2 supercomputer hosted at the Supercomputing Laboratory at KAUST