Preprints
https://doi.org/10.5194/gmd-2020-325
https://doi.org/10.5194/gmd-2020-325

Submitted as: development and technical paper 06 Nov 2020

Submitted as: development and technical paper | 06 Nov 2020

Review status: a revised version of this preprint is currently under review for the journal GMD.

Lossy Checkpoint Compression in Full Waveform Inversion

Navjot Kukreja1, Jan Hückelheim2, Mathias Louboutin3, John Washbourne4, Paul H. J. Kelly5, and Gerard J. Gorman1 Navjot Kukreja et al.
  • 1Department of Earth Science and Engineering, Imperial College London
  • 2Argonne National Laboratory
  • 3Georgia Institute of Technology
  • 4Chevron Corporation
  • 5Department of Computing, Imperial College London

Abstract. This paper proposes a new method that combines check- pointing methods with error-controlled lossy compression for large-scale high-performance Full-Waveform Inversion (FWI), an inverse problem commonly used in geophysical exploration. This combination can signif- icantly reduce data movement, allowing a reduction in run time as well as peak memory.

In the Exascale computing era, frequent data transfer (e.g., memory bandwidth, PCIe bandwidth for GPUs, or network) is the performance bottleneck rather than the peak FLOPS of the processing unit.

Like many other adjoint-based optimization problems, FWI is costly in terms of the number of floating-point operations, large memory foot- print during backpropagation, and data transfer overheads. Past work for adjoint methods has developed checkpointing methods that reduce the peak memory requirements during backpropagation at the cost of additional floating-point computations.

Combining this traditional checkpointing with error-controlled lossy compression, we explore the three-way tradeoff between memory, precision, and time to solution. We investigate how approximation errors introduced by lossy compression of the forward solution impact the objective function gradient and final inverted solution. Empirical results from these numerical experiments indicate that high lossy-compression rates (compression factors ranging up to 100) have a relatively minor impact on convergence rates and the quality of the final solution.

Navjot Kukreja et al.

 
Status: final response (author comments only)
Status: final response (author comments only)
AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment
[Login for authors/topical editors] [Subscribe to comment alert] Printer-friendly Version - Printer-friendly version Supplement - Supplement

Navjot Kukreja et al.

Data sets

SEG/EAGE 3-D Salt and Overthrust Models F. Aminzadeh, J. Brac, and T. Kunz https://doi.org/10.1190/1.1437283

Model code and software

Lossy Checkpoint Compression in Full Waveform Inversion (accompanying code) N. Kukreja and M. Louboutin https://doi.org/10.5281/zenodo.4247198

Navjot Kukreja et al.

Viewed

Total article views: 657 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
486 167 4 657 11 7
  • HTML: 486
  • PDF: 167
  • XML: 4
  • Total: 657
  • BibTeX: 11
  • EndNote: 7
Views and downloads (calculated since 06 Nov 2020)
Cumulative views and downloads (calculated since 06 Nov 2020)

Viewed (geographical distribution)

Total article views: 486 (including HTML, PDF, and XML) Thereof 482 with geography defined and 4 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 

Cited

Latest update: 26 Oct 2021
Download
Short summary
Full Waveform Inversion is a PDE-constrained optimization problem that is notorious for its high computational load and memory footprint. In this paper we present a method that combines recomputation with lossy compression to accelerate the computation with minimal loss of precision in the results. We show this using experiments running FWI with a variety of compression settings on a popular academic dataset.