RTM Solver

Accelerated RTM SolversSoftware for accelerated processing of RTM

Reverse Time Migration (RTM) is the current ‘state-of-the-art’ in seismic imaging. The strength of RTM stems from the fact that it fully respects the two-way acoustic wave equation, thus improving imaging in areas where complex geology violates the assumptions made in Kirchhoff or one-way wave equation migrations. Until recently, RTM’s widespread use was severely hindered by the enormous computing resources required to process the data. This computational bottleneck is now cleared with Acceleware’s patent-pending software solution AxRTM™.

AxRTM provides the core numerical functionality of Reverse Time Migration as a library that can be integrated into an existing seismic processing framework. AxRTM has a modular architecture supporting a variety of integrator-supplied functionality, and currently supports both optimized multi-core CPU and NVIDIA GPU hardware.

Acceleware RTM Layer Model

The image above illustrates how Acceleware's RTM multi-core libraries fit into your workflow. Paradigm and Tsunami also offer integrated solutions. Find out more on our Oil & Gas Solutions page.

The Importance of Anisotropic Reverse Time Migration

Move the mouse over the picture to show how isotropic imaging of this anisotropic dataset fails to image the salt flanks and mis-positions the sub-salt sediments. In the TTI image, the salt flanks are clearly defined and the sediments imaged right up to the salt boundary.

Features

The table below shows major components of the AxRTM library. The architecture of AxRTM is designed to allow for the addition of geophysical functionality in a modular fashion. This allows an Integrator who has their own RTM expertise to incorporate their proprietary methods above and/or into the API.

  • Forward/Back Propagators
  • Optimized CPU and GPU
  • Finite-difference
    • 4th, 8th order spatial; 2nd order time
    • Other orders easily supported
  • Isotropic acoustic
  • Anisotropic
    • VTI (Vertical Transverse Isotropy)
    • TTI (Tilted Transverse Isotropy)
  • Pseudo-spectral Domain
  • Domain Decomposition
  • Support for multi-core, multi-socket and NUMA (non-uniform memory access vis a vis Nehalem and Opteron)
  • Support for multi-GPU attached to the same compute node (up to 4 GPU’s, no theoretical limit)
  • Support for distributed memory clusters using MPI (OpenMPI) or sockets to connect nodes; both accelerated GPU nodes or traditional CPU-only server nodes
  • Imaging Condition
  • Simple cross-correlation
  • High quality imaging condition (removes self-correlation noise in final image)
  • Illumination Calculation
  • (Optional) source illumination calculation performed during the forward pass
  • Illumination information can be used for amplitude compensation of the final migrated image volume
  • Boundary Conditions
  • Sponge (Israeli-Orszag)
  • Perfectly matched layer, with VTI termination
  • Perfectly reflective
  • Wavefield History
  • High performance low storage requirements without loss of fidelity

Currently supported hardware is Intel/AMD multi-core CPU and NVIDIA GPUs (Tesla-8 and Tesla-10 or equivalents). Support for future computational hardware was part of the fundamental design of the AxRTM library, meaning that adopting a new hardware platform in the future will not require any changes to the overlying RTM application.

Domain Decomposition

For large RTM volumes, AxRTM decomposes the domain across multiple compute nodes. Acceleware fully supports the spanning of a single migration volume across both multiple GPUs/CPUs on a single compute node, as well as multiple compute nodes (Figure 1). Acceleware has migrated domains up to 16 Tesla-S nodes leading to a maximum deployed size of 256GB, for a single accelerated RTM domain/shot migration, using Tesla-10 series hardware (4GB per GPU; 64 GPUs). Although there is no limit to the number of nodes in the software.

Figure 1: Compute nodes are dynamically grouped into Shot Groups to work on very large migration volumes. Each Tesla S1070 in this Group has 4x 4GB of RAM for a Shot Group total of 48GB.