Devito: Automated Fast Finite Difference Computation

Kukreja, Navjot; Louboutin, Mathias; Vieira, Felippe; Luporini, Fabio; Lange, Michael; Gorman, Gerard J.

doi:10.1109/wolfhpc.2016.06

Cited by 24 publications

(18 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We perform numerical solution of the acoustic waveequation and the respective adjoint computation using the domain-specific symbolic language Devito (Kukreja et al, 2016;Louboutin et al, 2017). Numerical solution is performed using a fourth-order finite-difference scheme.…”

Section: Adjoint-state Methodsmentioning

confidence: 99%

Stochastic Seismic Waveform Inversion Using Generative Adversarial Networks as a Geological Prior

2019

View full text Add to dashboard Cite

We present an application of deep generative models in the context of partial-differential equation (PDE) constrained inverse problems. We combine a generative adversarial network (GAN) representing an a priori model that creates subsurface geological structures and their petrophysical properties, with the numerical solution of the PDE governing the propagation of acoustic waves within the earth's interior. We perform Bayesian inversion using an approximate Metropolis-adjusted Langevin algorithm (MALA) to sample from the posterior given seismic observations. Gradients with respect to the model parameters governing the forward problem are obtained by solving the adjoint of the acoustic wave-equation. Gradients of the mismatch with respect to the latent variables are obtained by leveraging the differentiable nature of the deep neural network used to represent the generative model. We show that approximate MALA sampling allows efficient Bayesian inversion of model parameters obtained from a prior represented by a deep generative model, obtaining a diverse set of realizations that reflect the observed seismic response.

show abstract

Section: Adjoint-state Methodsmentioning

confidence: 99%

Stochastic Seismic Waveform Inversion Using Generative Adversarial Networks as a Geological Prior

2019

View full text Add to dashboard Cite

show abstract

“…Para obter o máximo desempenho nesses novos dispositivos,é necessária alguma reengenharia das regiões do código, se não toda a aplicação. Assim, [Kukreja et al 2016] gera automaticamente um código para simulação geofísica altamente otimizado para múltiplas arquiteturas, enquanto [Niu et al 2014] sugere o uso de reconfiguração de sistemas em tempo de execução e um modelo de desempenho, para reduzir o consumo de recursos computacionais. [Caballero et al 2015] estudou o efeito de diferentes otimizações em nível de memória e computação em equações de propagação de ondas elásticas.…”

Section: Trabalhos Relacionadosunclassified

Portabilidade e Eficiência do Método Fletcher de Aplicações Sísmicas em Arquiteturas Multicore e GPU

Serpa

Pavan

Panetta

et al. 2019

Anais Do XX Simpósio Em Sistemas Computacionais De Alto Desempenho (SSCAD 2019)

View full text Add to dashboard Cite

A simulação da propagação de ondas acústicas é a base das ferramentas de imagem sı́smica utilizadas pela indústria de petróleo e gás. Para realizar tais simulações, arquiteturas de CAD são empregadas, fornecendo resultados mais rápidos e com maior precisão a cada geração de processadores. Entretanto, para atingir alto desempenho nessas arquiteturas, vários desafios devem ser levados em consideração no momento do desenvolvimento da aplicação. Neste artigo, a Modelagem Fletcher foi otimizada para multicore e GPU e o desempenho, o consumo de energia e a eficiência energética de oito versões do código foram avaliados. Os resultados mostram que a versão CUDA tem o melhor desempenho e eficiência energética; no entanto, a versão OpenACC que tem a vantagem da portabilidade, tem um desempenho e degradação de eficiência energética de apenas 10 e 8% comparado com CUDA. ∗

show abstract

“…The adoption of such an abstraction is advantageous since it removes the need for the user to expand the equations by hand which can be an error-prone task. Furthermore, much like the Devito domain specific language (DSL) [20,21] for finite difference stencil compilation, OpenSBLI makes use of the SymPy symbolic algebra library that supplies the basic components required for the modelling functionality that has been implemented in the present work. This functionality includes the automatic expansion of indices based on their contraction structure, such that repeated indices are expanded into a sum about that index, and the implementation of various types of differential operator.…”

Section: Equation Specificationmentioning

confidence: 99%

OpenSBLI: A framework for the automated derivation and parallel execution of finite difference solvers on a range of computer architectures

Jacobs

Jammy

Sandham

2017

Journal of Computational Science

View full text Add to dashboard Cite

Exascale computing will feature novel and potentially disruptive hardware architectures. Exploiting these to their full potential is non-trivial. Numerical modelling frameworks involving finite difference methods are currently limited by the 'static' nature of the hand-coded discretisation schemes and repeatedly may have to be re-written to run efficiently on new hardware. In contrast, OpenSBLI uses code generation to derive the model's code from a high-level specification. Users focus on the equations to solve, whilst not concerning themselves with the detailed implementation. Source-to-source translation is used to tailor the code and enable its execution on a variety of hardware.

show abstract

Devito: Automated Fast Finite Difference Computation

Cited by 24 publications

References 26 publications

Stochastic Seismic Waveform Inversion Using Generative Adversarial Networks as a Geological Prior

Stochastic Seismic Waveform Inversion Using Generative Adversarial Networks as a Geological Prior

Portabilidade e Eficiência do Método Fletcher de Aplicações Sísmicas em Arquiteturas Multicore e GPU

OpenSBLI: A framework for the automated derivation and parallel execution of finite difference solvers on a range of computer architectures

Contact Info

Product

Resources

About