Accelerating Seismic Redatuming Using Tile Low-Rank Approximations on NEC SX-Aurora TSUBASA

Hong, Yuxi; Ltaief, Hatem; Ravasi, Matteo; Gatineau, Laurent; Keyes, David E.

doi:10.14529/jsfi210201

Cited by 7 publications

(15 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Whilst in Hong et al (2021), we focused only on the implementation of the forward batch MVM operation; here, we cover the entire SRI workflow where an iterative solver is used to solve the Marchenko equations. This requires the application of four batch MVM operations as explained in Section 4.…”

Section: Impact On Sri Computational Stagesmentioning

confidence: 99%

High performance computing seismic redatuming by inversion with algebraic compression and multiple precisions

Hong,

Ltaief,

Ravasi

et al. 2024

The International Journal of High Performance Computing Applica

View full text Add to dashboard Cite

We present a high-performance implementation of Seismic Redatuming by Inversion (SRI), which combines algebraic compression with mixed-precision (MP) computations. Seismic redatuming entails the repositioning of seismic data recorded at the surface of the Earth to a subsurface level closer to where reflections have originated. Marchenko-based redatuming is the best-in-class SRI method from a theoretical point of view because it can accurately handle full seismic wavefields with minimal data pre-processing. However, it requires intensive data movement, a large memory footprint, and extensive computations due to the need for manipulating large, formally dense, complex-valued frequency matrices in the bandwidth of the recorded seismic data. More precisely, at the heart of Marchenko modeling operator lies a Multi-Dimensional Convolution (MDC) operator, which requires the application of repeated complex-valued Matrix-Vector Multiplications (MVMs) with chosen set of seismic frequency matrices. SRI is deemed to be too time-consuming for realistic domain sizes and therefore industrially impractical. To alleviate these computational bottlenecks, we first improve the memory footprint of the MDC operator by using Tile Low-Rank Matrix-Vector Multiplications (TLR-MVMs); this exploits the low-rank nature of seismic recordings in the frequency domain, used as the kernel of the MDC modeling operator. MP computations are further introduced to increase the arithmetic intensity of the operators. We validate the numerical robustness of our synergistic approach on a benchmark 3D synthetic seismic dataset and demonstrate its use on various hardware architectures. We develop several FP32/FP16/INT8 MP TLR-MVM implementations with a limited and controllable impact on the overall accuracy of application. We achieve up to 63X memory footprint reduction and 12.5X performance speedup against the traditional dense MVM kernel on a single NVIDIA A100 GPU. Moreover, linear scalability is attained when deploying our implementation on eight A100 GPUs. Our implementations of the MDC operator integrate with the open-source PyLops framework for large-scale, matrix-free inverse problems.

show abstract

Section: Impact On Sri Computational Stagesmentioning

confidence: 99%

High performance computing seismic redatuming by inversion with algebraic compression and multiple precisions

Hong,

Ltaief,

Ravasi

et al. 2024

The International Journal of High Performance Computing Applica

View full text Add to dashboard Cite

show abstract

“…However, whilst this seems to be of vital importance for the former method, the improvement is fairly minor in the latter, making it possible to work with a single or small group of virtual sources at the same time without seriously compromising the quality of the reconstruction. This finding is of great importance for the application of MDD to large-scale, three dimensional datasets where solving for the entire set of virtual sources at the same time may be beyond reach of our current compute capabilities [27], [34]. In order to be able to assess the performance of stochastic MDD against full-gradient MDD, we begin by creating a synthetic, Volve-like dataset.…”

Section: B Synthetic Subsalt Redatumingmentioning

confidence: 99%

Stochastic Multi-Dimensional Deconvolution

Ravasi

Selvan

Luiken

2022

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

Geophysical measurements such as seismic datasets contain valuable information that originate from areas of interest in the subsurface; these seismic reflections are however inevitably contaminated by other events created by waves reverberating in the overburden. Multi-Dimensional Deconvolution (MDD) is a powerful technique used at various stages of the seismic processing sequence to create ideal datasets deprived of such overburden effects. Whilst the underlying forward problem holds for a single source, a successful inversion of the MDD equations requires availability of a large number of sources alongside prior information, possibly introduced in the form of physical constraints (e.g., reciprocity and causality). In this work, we present a novel formulation of time-domain MDD based on a finite-sum functional. The associated inverse problem is then solved by means of stochastic gradient descent algorithms, where the gradients at each iteration are computed using a small subset of randomly selected sources. Through synthetic and field data examples, we show that the proposed method converges more stably than the conventional approach based on full gradients. Stochastic MDD represents a novel, efficient, and robust strategy to deconvolve seismic wavefields in a multi-dimensional fashion.

show abstract

“…Once such bases are available, the standard batched MVM step in the MDC operator is replaced by a batched TLR-MVM algorithm (Ltaief et al, 2021;Hong et al, 2021). For each frequency, this is performed in three consecutive phases: (1) the…”

Section: Tile Low-rank Compressionmentioning

confidence: 99%

“…To alleviate the 'curse of dimensionality', these authors showed that by factorizing the kernel matrices of the MDC operator by means of Truncated Singular Value Decomposition (TSVD), they could reduce both the memory footprint and computational cost of applying such an operator in a 2D scenario. Following their claim that similar benefits could be obtained when working with 3D datasets, Hong et al (2021) proposed to apply the so-called Tile Low-Rank (TLR) compression algorithm to the frequency slices of a synthetic 3D dataset.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Large-scale Marchenko imaging with distance-aware matrix reordering, tile low-rank compression, and mixed-precision computations

Ravasi

Hong

Ltaief

et al. 2022

Second International Meeting for Applied Geoscience &Amp; Energy

View full text Add to dashboard Cite

A variety of wave-equation-based seismic processing algorithms rely on the repeated application of the Multi-Dimensional Convolution (MDC) operator. For large-scale 3D seismic surveys, this comes with severe computational challenges due to the sheer size of high-density, full-azimuth seismic datasets required by such algorithms. We present a three-fold solution that greatly alleviates the memory footprint and computational cost of 3D MDC by leveraging a combination of i) distance-aware matrix reordering, ii) Tile Low-Rank (TLR) matrix compression, and iii) computations in mixed floating-point precision. By applying our strategy to a 3D synthetic dataset, we show that the size of kernel matrices used in the Marchenko redatuming and Multi-Dimensional Deconvolution equations can be reduced by a factor of 34 and 6, respectively. We also introduce a TLR Matrix-Vector Multiplication (TLR-MVM) algorithm that, as a direct consequence of such compression capabilities, is consistently faster than its dense counterpart by a factor of 4.8 to 36.1 (depending on the selected hardware). As a result, the associated inverse problems can be solved at a fraction of cost in comparison to state-ofthe-art implementations that require a pass through the entire data at each MDC operation. This is achieved with minimal impact on the quality of the processing outcome.

show abstract

Accelerating Seismic Redatuming Using Tile Low-Rank Approximations on NEC SX-Aurora TSUBASA

Cited by 7 publications

References 34 publications

High performance computing seismic redatuming by inversion with algebraic compression and multiple precisions

High performance computing seismic redatuming by inversion with algebraic compression and multiple precisions

Stochastic Multi-Dimensional Deconvolution

Large-scale Marchenko imaging with distance-aware matrix reordering, tile low-rank compression, and mixed-precision computations

Contact Info

Product

Resources

About