Time filter

Source Type

Lemieux J.-F.,McGill University | Tremblay B.,McGill University | Sedlacek J.,ETH Zurich | Tupper P.,Simon Fraser University | And 3 more authors.
Journal of Computational Physics | Year: 2010

We have implemented the Jacobian-free Newton-Krylov (JFNK) method to solve the sea ice momentum equation with a viscous-plastic (VP) formulation. The JFNK method has many advantages: the system matrix (the Jacobian) does not need to be formed and stored, the method is parallelizable and the convergence can be nearly quadratic in the vicinity of the solution. The convergence rate of our JFNK implementation is characterized by two phases: an initial phase with slow convergence and a fast phase for which the residual norm decreases significantly from one Newton iteration to the next. Because of this fast phase, the computational gain of the JFNK method over the standard solver used in existing VP models increases with the required drop in the residual norm (termination criterion). The JFNK method is between 3 and 6.6 times faster (depending on the spatial resolution and termination criterion) than the standard solver using a preconditioned generalized minimum residual method. Resolutions tested in this study are 80, 40, 20 and 10 km. For a large required drop in the residual norm, both JFNK and standard solvers sometimes do not converge. The failure rate for both solvers increases as the grid is refined but stays relatively small (less than 2.3% of failures). With increasing spatial resolution, the velocity gradients (sea ice deformations) get more and more important. Nonlinear solvers such as the JFNK method tend to have difficulties when there are such sharp structures in the solution. This lack of robustness of both solvers is however a debatable problem as it mostly occurs for large required drops in the residual norm. Furthermore, when it occurs, it usually affects only a few grid cells, i.e., the residual is small for all the velocity components except in very localized regions. Globalization approaches for the JFNK solver, such as the line search method, have not yet proven to be successful. Further investigation is needed. © 2009 Elsevier Inc. All rights reserved.


Weldon M.,Acceleware Corporation | Maxwell L.,Acceleware Corporation | Cyca D.,Acceleware Corporation | Hughes M.,Acceleware Corporation | And 3 more authors.
Applied Computational Electromagnetics Society Journal | Year: 2010

This paper outlines several key features and conditions that impact the performance of FDTD on GPUs. It includes relevant performance measurements as well as practical suggestions on how to mitigate their impact. Among these factors are: PML depth, the number of unique materials, dispersive materials, the impact of field reads/observations, simulation orientation, and domain decomposition using multiple GPUs. The paper shows that the performance of FDTD on GPUs can be limited in certain extreme cases, but with proper care on the part of the designer these cases can be managed and maximum performance guaranteed. © 2010 ACES.


Ong C.Y.,Acceleware Corporation | Weldon M.,Acceleware Corporation | Quiring S.,Acceleware Corporation | Maxwell L.,Acceleware Corporation | And 3 more authors.
IEEE Microwave Magazine | Year: 2010

Electromagnetic (EM) simulators are essential tools in the design cycle of today's complex micro wave systems. Accurate EM simulations allow designers to gain better understanding of their designs, make more informed decisions, and, consequently, produce higher quality designs. In a competitive environment, where time-to-market is critical, acceleration of the design simulation cycle is an enabler for companies looking to capture market share and revenue. © 2010 IEEE.


Pasalic D.,Acceleware Ltd | Vaca P.,Acceleware Ltd | Okoniewski M.,University of Calgary
Proceedings - 2014 International Conference on Electromagnetics in Advanced Applications, ICEAA 2014 | Year: 2014

An algorithm for rigorous analysis of electromagnetic (EM) heating of heavy oil reservoirs is presented in the paper. The algorithm combines an FDTD-based EM solver with a reservoir simulator, utilizing the fact that the two simulators operate at completely different time scales. The two simulators also utilize different grids. Special attention needs to be taken when interpolating the values between the two grids to avoid occurrence of non-physical effects. An example of a vertical 50-m long dipole antenna heating a heavy oil reservoir in northern Alberta, Canada is presented. © 2014 IEEE.


Qin Y.,Acceleware Ltd | Okoniewski M.,Acceleware Ltd
76th European Association of Geoscientists and Engineers Conference and Exhibition 2014: Experience the Energy - Incorporating SPE EUROPEC 2014 | Year: 2014

Reverse-Time Migration (RTM) is routinely used for depth velocity model building and imaging in deepwater exploration around the world. In this study, we employed analytical redatuming to improve the efficiency of true-amplitude RTM for deepwater Earth model. This method analytically redatums both the source and receiver wavefields from surface to water bottom and injects the redatumed wavefield into the computational domain while preserving true-amplitude in the migrated common-shot image. By using analytic redatuming in deepwater area, the total runtime, memory and disk space requirements are significantly reduced. In addition, the image quality is improved by overcoming the singularity of pointsource injection in the finite difference method, accurate positioning of source and receivers, and the reduction of numerical dispersion noise. The underlying theories developed in this paper can be extended to some other important applications such as Full-Waveform Inversion (FWI), true-amplitude layerstripping RTM/FWI.


Clark G.,Acceleware Ltd
2nd EAGE Workshop on High Performance Computing for Upstream | Year: 2015

Full waveform inversion (FWI) is an increasingly popular algorithm for automatically improving earth models in seismic exploration. The method is incredibly computationally expensive because the earth model is built up with many iterations of forward modelling and reverse time migration (RTM). Current research is helping to reduce the number of iterations and improve the robustness of convergence, but the prohibitive cost of FWI makes running real world datasets impractical for many researchers. Furthermore, information regarding the geological region can be used to accelerate convergence. Therefore, an FWI implementation must be both high performance, and flexible. This commercial case study outlines our approach to minimizing the cost of a flexible implementation. The key feature of our approach is that we divide the algorithm into low cost and high cost steps. Fortunately, the low cost steps in the FWI algorithm are also the ones subject to the most research. The extremely high cost full wave modelling steps are comparatively consistent between variations of the FWI algorithm. Therefore, the full wave modelling is hardware optimized and all other steps can be quickly rewritten in Python. Copyright © 2015 by the European Association of Geoscientists & Engineers (EAGE) All rights reserved.


McGarry R.,Acceleware Corporation | Pasalic D.,Acceleware Corporation | Ong C.,Acceleware Corporation
SEG Technical Program Expanded Abstracts | Year: 2011

We present a scheme for 2D and 3D modeling of elastic wave propagation in general anisotropic media using finite difference approximations on a Lebedev grid. Key to the efficiency and practicality of such a scheme is minimization of the number of spatial points at which wavefield components are calculated and stored. To this end we introduce a method for calculating dispersion-reducing finite difference coefficients. We also demonstrate a very useful property of Lebedev grids, namely grid decoupling. Finally, we present a numerical example showing the effectiveness of our proposed scheme by comparison with an analytical solution. © 2011 Society of Exploration Geophysicists.


Ong C.,Acceleware Corporation | Pasalic D.,Acceleware Corporation | McGarry R.,Acceleware Corporation
SEG Technical Program Expanded Abstracts | Year: 2011

A common finite difference implementation of reverse time migration with tilted transverse isotropy (TTI) follows the formulation of Alkhalifah (2000), Fletcher et. al. (2008) and Zhou et. al. (2006). The finite difference implementation of these equations in Cartesian coordinates necessitates the computation of mixed partial derivatives of the acoustic pressure in the spatial domain. Different methods exist for the computation of these mixed derivatives including sequentially computing centered first derivatives in each direction, computing staggered first derivatives with interpolation or the pseudospectral method. The computation of centered first derivatives causes ringing in the output but using staggered derivatives requires interpolation of the staggered points back to the original grid. The pseudospectral method necessitates the computation of a 2D FFT in a 2D simulation. In this paper, we propose a hexagonal mesh for the finite difference implementation of 2D TTI RTM. The implementation eliminates the need for mixed partial derivatives and reduces grid dispersion in wave simulations. © 2011 Society of Exploration Geophysicists.


Pasalic D.,Acceleware Ltd | Okoniewski M.,Acceleware Ltd | Okoniewski M.,University of Calgary
Proceedings of the 2012 International Conference on Electromagnetics in Advanced Applications, ICEAA'12 | Year: 2012

This paper presents a fast implementation of Controlled Source Electromagnetics (CSEM) method. The speed of the method results from algorithmic improvements inherent to the Mittet's method [4], as well as modifications and to air-water boundary conditions and the transforms used. All compute intensive aspects of the method were implemented on graphical processing units processors (GPU). © 2012 IEEE.


Nauta M.D.,Acceleware Ltd | Cyca D.L.,Acceleware Ltd
EAGE Workshop on High Performance Computing for Upstream 2014 | Year: 2014

This paper examines the computational trade-offs of higher order in time for Reverse Time Migration (RTM). It compares the hardware trade-offs for running RTM with 2nd order accurate temporal derivatives and 4th order accurate temporal derivatives. The exact difference in computational complexity depends on the RTM implementation. The timestep using 4th order accuracy can be much longer for the same level of accuracy theoretically leading to an overall savings in computational resources. This analysis is true for the main propagation, however practical RTM is often limited by hard disk speed, available random access memory (RAM), or data transfers between compute nodes in a distributed system. This paper discusses these additional elements and how they impact performance.

Loading Acceleware Corporation collaborators
Loading Acceleware Corporation collaborators