Implementation and measurements of an efficient Fixed Point Adjoint

Abstract : Efficient Algorithmic Differentiation of Fixed-Point loops requires a specific strategy to avoid explosion of memory requirements. Among the strategies documented in literature, we have selected the one introduced by B. Christianson. This method features original mechanisms such as repeated access to the trajectory stack or duplicated differentiation of the loop body with respect to different independent variables. We describe in this paper how the method must be further specified to take into account the particularities of real codes, and how data flow information can be used to automate detection of relevant sets of variables. We describe the way we implement this method inside an AD tool. Experiments on a medium-size application demonstrate a minor, but non negligible improvement of the accuracy of the result, and more importantly a major reduction of the memory needed to store the trajectories.
Complete list of metadatas

Cited literature [7 references]  Display  Hide  Download

https://hal.inria.fr/hal-01244298
Contributor : Laurent Hascoet <>
Submitted on : Tuesday, December 15, 2015 - 3:51:35 PM
Last modification on : Tuesday, January 30, 2018 - 4:56:01 PM
Long-term archiving on : Saturday, April 29, 2017 - 3:10:34 PM

File

FixedPointImplementationPaper....
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01244298, version 1

Collections

Citation

Ala Taftaf, Laurent Hascoët, Valérie Pascual. Implementation and measurements of an efficient Fixed Point Adjoint. EUROGEN 2015, ECCOMAS, Sep 2015, GLASGOW, United Kingdom. ⟨hal-01244298⟩

Share

Metrics

Record views

169

Files downloads

89