Guillaume Aupy (Vanderbilt Univ.) will give a talk on Optimal Multi-stage algorithm for Adjoint Computation.
Abstract: We reexamine the work of Stumm and Walther on multistage algorithms for adjoint
computation. We provide an optimal algorithm for this problem whenthere are two levels of check- points, in memory and on disk. Previously, optimal algorithms for adjoint computations were known only for a single level of checkpoints with no writing and rea ding costs; a well-known example is the binomial
checkpointing algorithm of Griewank and Walth er. Stumm and Walther extended that binomial checkpointing algorithm to the case of two levels of checkpoints, but they did not provide any optimality results. We bridge the gap by designing the first optimal algorithm in this context. We experimentally
compare our optimal algorithm with that o f Stumm and Walther to assess the difference in performance.