Where: Houston, Texas, USA
Date: September 22-27, 2013
Stop by the Acceleware booth at SEG to find out the latest in seismic exploration and high performance computing. Learn about the latest algorithm developments for FWI and 3D Elastic modeling, hardware acceleration advancements and product updates for Acceleware’s Reverse Time Migration (RTM) library and forward modeling solver for marine CSEM.
True-amplitude common-shot acoustic reverse time migration (RTM)
Presented by: Yilong Qin, Acceleware Ltd.
Thu September 26 @9:45am, SPMI 6: Advanced Migration Algorithms Session
Generating true-amplitude acoustic RTM shot images is an important step for improving the quality of an RTM image. One way to generate a true-amplitude RTM common-shot image is via Least-Squares RTM (LSRTM). But iterative LSRTM can be computationally expensive. In practice, a cross-correlation imaging condition (CCIC) with source illumination normalization is often used in the conventional time-domain RTM. However, the CCIC with source illumination normalization does not produce a true-amplitude RTM common-shot image, if the directional fold correction factor (or the horizontal slowness at each receiver) is not accounted for while back-propagating the receiver wavefield.
In this presentation, enlightened by the true-amplitude Kirchhoff common-shot inversion formula, we propose an approach to generate the true-amplitude RTM common-shot image by incorporating the directional fold correction into the receiver wavefield. Synthetic examples in constant and v(z) models show that, in the absence of low-frequency self-correlation noises, a true-amplitude acoustic RTM common-shot image can be obtained by combining the source-illumination-normalized cross-correlation imaging condition with the directional fold correction factor. Imaging of a modified version of the Marmousi-II model is also shown to demonstrate the effectiveness of our proposed approach.
Direction-vector-based angle gathers from anisotropic elastic RTM
Presented by: Ray McGarry, Acceleware Ltd.
Wed September 25 @10:35am, SPMI 4: Anisotropy, Elasticity, and Attenuation Session
In this session we will present a method for calculating angle gathers from anisotropic elastic RTM. Reflection angle and azimuth are determined from direction vectors calculated directly from the propagating wavefields. We show that, depending on how the direction vectors are calculated, it may be necessary to convert from group angle to phase angle as part of the angle gather binning process. We give a simple expression for performing this conversion in the case of qPwave components. Finally, we show that our method is capable of accurate reproduction of theoretical PP reflection coefficients for anisotropic elastic models.
Optimizing RTM for Intel® Xeon Phi™ Coprocessors: Lessons Learned
Presented by: Darren Foltinek, Acceleware
When: Wednesday, September 25 at 10:00am
Where: High Performance Computing Theater (exhibition floor #3501)
In this presentation we will share our experience in porting and optimizing a commercial Reverse Time Migration library to a cluster of Xeon Phi Coprocessors. We first highlight the similarities of optimizing a TTI propagation kernel for Xeon Phi and Xeon Sandy Bridge, demonstrating how we optimized the OpenMP pragmas and tuned the auto-vectorizer. The second part of the presentation will focus on achieving linear scaling across multiple devices and nodes, which is critical when dealing with real-world problem sizes. We will discuss how to multi-task communication between Xeon Phi devices, hard disk IO, and the core computation kernel. Finally, we will provide an overview of the Linux set up and the additional tools that were used.
An Introduction to Full Waveform Inversion using the AxRTM™ Library
Presented by Dr. Ray McGarry, Acceleware
Mon September 23 @11:00am, Tue September 24 @2:00pm, Wed September 25 @2:00pm
Full Waveform Inversion (FWI) is a technology whose time has come. In this presentation Acceleware will introduce the basic concepts of FWI. We will show, through examples, that FWI is very closely related to the Reverse Time Migration (RTM) method of seismic imaging and demonstrate how the Acceleware AxRTM library may be used to invert seismic data to derive high quality earth models of unprecedented detail.
AxRTM: An Industry Leading Reverse Time Migration Library
Presented by Darren Foltinek, Acceleware
Mon September 23 @2:00pm, Tue September 24 @11:00am, Wed September 25 @1:00pm
AxRTM is Acceleware's CPU & GPU enabled Reverse Time Migration library. This presentation will profile recent product enhancements including: low-cost and effective removal of the RTM low-frequency noise, improved amplitude correctness for better image quality in shallow sediments, improved speed for TTI, VTI and isotropic media and support for the latest NVIDIA® Tesla GPUs and Intel® Xeon Phi™ coprocessors. Performance figures and image results will be examined.
How Systems and Software Must Evolve to Satisfy Seismic Imaging Challenges
Presented by Ty McKercher, NVIDIA
Tue September 24 @10:00am, Wed September 25 @10:00am
This presentation will cover how GPUs have evolved to solve complex seismic imaging challenges. Over the past five years, three generations of NVIDIA GPUs have delivered affordable seismic imaging techniques, including RTM, FWI, and Elastic algorithms. During this evolution, earth models have significantly improved, yielding better drilling decisions. Customer performance proof points will be used to highlight real world experiences.
High Performance Computing Platforms for Geophysics in the Coming Decade
Presented by Philippe Thierry, Intel
Mon September 23 @1:00pm, Tue September 24 @1:00pm
Can seismic imaging and inversion problems be scaled to the next generation super computer as data volume and complexity continue to grow? This presentation will examine the compute demands of the industry and how technology may meet this challenge. An overview of Intel’s current and future hardware will be provided as well as planned software advancements.
RTM Product Manager, Acceleware Ltd.
Dr Ray McGarry
Seismic Research Team Lead, Acceleware Ltd.
Dr. Yilong Qin
Geophysical Researcher, Acceleware Ltd.
Principal Solution Architect, NVIDIA
Energy Engineering Team Manager, Intel