Skip to Main content Skip to Navigation
Journal articles

Adjoint computations by algorithmic differentiation of a parallel solver for time-dependent PDEs

Abstract : A computational fluid dynamics code is differentiated using algorithmic differentiation (AD) in both tangent and adjoint modes. The two novelties of the present approach are (1) the adjoint code is obtained by letting the AD tool Tapenade invert the complete layer of message passing interface (MPI) communications, and (2) the adjoint code integrates time-dependent, non-linear and dissipative (hence physically irreversible) PDEs with an explicit time integration loop running for ca. 106 time steps. The approach relies on using the Adjoinable MPI library to reverse the non-blocking communication patterns in the original code, and by controlling the memory overhead induced by the time-stepping loop with binomial checkpointing. A description of the necessary code modifications is provided along with the validation of the computed derivatives and a performance comparison of the tangent and adjoint codes.
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-03033123
Contributor : Open Archive Toulouse Archive Ouverte (oatao) <>
Submitted on : Tuesday, December 1, 2020 - 11:14:30 AM
Last modification on : Friday, January 8, 2021 - 10:46:02 AM

File

Cardesa_26554.pdf
Files produced by the author(s)

Identifiers

Collections

Citation

José Cardesa, Laurent Hascoët, Christophe Airiau. Adjoint computations by algorithmic differentiation of a parallel solver for time-dependent PDEs. Journal of computational science, Elsevier, 2020, 45, pp.101155. ⟨10.1016/j.jocs.2020.101155⟩. ⟨hal-03033123⟩

Share

Metrics

Record views

10

Files downloads

20