Submitted as: model description paper 05 Aug 2021

Submitted as: model description paper | 05 Aug 2021

Review status: this preprint is currently under review for the journal GMD.

An explicit GPU-based material point method solver for elastoplastic problems (ep2-3De v1.0)

Emmanuel Wyser1,2, Yury Alkhimenkov1,2,3, Michel Jaboyedoff1,2, and Yury Y. Podladchikov1,2,3 Emmanuel Wyser et al.
  • 1Institute of Earth Sciences, University of Lausanne, 1015 Lausanne, Switzerland
  • 2Swiss Geocomputing Center, University of Lausanne, 1015 Lausanne, Switzerland
  • 3Faculty of Mechanics and Mathematics, Lomonosov Moscow State University, Moscow, 119991, Russia

Abstract. We propose an explicit GPU-based solver within the material point method (MPM) framework on a single graphics pro- cessing unit (GPU) to resolve elastoplastic problems under two- and three-dimensional configurations (i.e., granular collapses and slumping mechanics). Modern GPU architectures, including Ampere, Turing and Volta, provide a computational framework that is well suited to the locality of the material point method in view of high-performance computing. For intense and nonlocal computational aspects (i.e., the back-and-forth mapping between the nodes of the background mesh and the material points), we use straightforward atomic operations (the scattering paradigm). We select the generalized interpolation material point method (GIMPM) to resolve the cell-crossing error, which typically arises in the original MPM, because of the C0 continuity of the linear basis function. We validate our GPU-based in-house solver by comparing numerical results for granular collapses with the available experimental data sets. Good agreement is found between the numerical results and experimental results for the free surface and failure surface. We further evaluate the performance of our GPU-based implementation for the three-dimensional elastoplastic slumping mechanics problem. We report i) a maximum performance gain of x200 between a CPU- and GPU-based implementation, provided that ii) the hardware limit (i.e., the peak memory bandwidth) of the device is reached. We finally showcase an application to slumping mechanics and demonstrate the importance of a three-dimensional configuration coupled with heterogeneous properties to resolve complex material behaviour.

Emmanuel Wyser et al.

Status: final response (author comments only)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on gmd-2021-200', Anonymous Referee #1, 20 Aug 2021
    • AC1: 'Reply on RC1', Emmanuel Wyser, 28 Aug 2021
  • RC2: 'Comment on gmd-2021-200', Quoc Anh Tran, 08 Sep 2021
    • AC2: 'Reply on RC2', Emmanuel Wyser, 23 Sep 2021

Emmanuel Wyser et al.

Model code and software

ep2-3De v1.0 Emmanuel Wyser, Yury Alkhimenkov, Michel Jaboyedoff and Yury Y. Podladchikov

Emmanuel Wyser et al.


Total article views: 507 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
396 101 10 507 2 1
  • HTML: 396
  • PDF: 101
  • XML: 10
  • Total: 507
  • BibTeX: 2
  • EndNote: 1
Views and downloads (calculated since 05 Aug 2021)
Cumulative views and downloads (calculated since 05 Aug 2021)

Viewed (geographical distribution)

Total article views: 466 (including HTML, PDF, and XML) Thereof 466 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 22 Oct 2021
Short summary
In this work, we propose an explicit implementation of the material point method under a graphical processing unit (GPU) architecture to solve for elastoplastic problems for three-dimensional configurations, such as the granular collapse or the slumping mechanics, i.e., landslide. The computational power of GPUs promotes fast code executions, compared to a traditional implementation under a central processing unit architecture. This allows the study of complex three-dimensional problems.