Reducing Memory Requirements in Scientific Computing and Optimal Control

摘要: In high accuracy numerical simulations and optimal control of time-dependent processes, often both many timesteps fine spatial discretizations are needed. Adjoint gradient computation, or post-processing simulation results, requires the storage solution trajectories over whole time, if necessary together with adaptively refined grids. this paper we discuss various techniques to reduce memory requirements, focusing first on data, which typically double precision floating point values. We highlight advantages disadvantages different approaches. Moreover, present an algorithm for efficient refined, hierarchic grids, integration compressed data.