GRASP with ejection chains for the dynamic memory allocation in embedded systems

Abstract

International audienceIn the design of electronic embedded systems, the allocation of data structures to memory banks is a main challenge faced by designers. Indeed, if this optimization problem is solved correctly, a great improvement in terms of efficiency can be obtained. In this paper, we consider the dynamic memory allocation problem, where data structures have to be assigned to memory banks in different time periods during the execution of the application. We propose a GRASP to obtain high quality solutions in short computational time, as required in this type of problem. Moreover, we also explore the adaptation of the ejection chain methodology, originally proposed in the context of tabu search, for improved outcomes. Our experiments with real and randomly generated instances show the superiority of the proposed methods compared to the state-of-the-art method

A multiple neighborhood search for dynamic memory allocation in embedded systems

“…We compare the MNS-TS with the iterative approaches (IA) , GRASP (Sevaux et al 2014) and the ILP formulation solved with Xpress-MP, that is used as a heuristic when the time limit of one hour is reached: the best solution found so far is then returned by the solver. Table 4 reports the main results of our computational tests.…”

“…Then, MNS-TS is compared with the ILP formulation, iterative approaches (IA) and GRASP (Sevaux et al 2014). Finally, we present the main results and statistical analyses to assess the performance of the MNS-TS approach.…”

“…where h(a, t, j) − h(a, t, x a,t ) is the solution cost variation resulting from moving data structure a from its current allocation x a,t to another memory bank j at time interval t. For the static memory allocation problem ) and the GRASP devised for the dynamic memory allocation (Sevaux et al 2014), we have used two neighborhoods: the first one consists of performing simple moves while satisfying the capacity of the memory banks. The second neighborhood is produced by ejection chains composed of an infeasible move (i.e., the capacity of a memory bank is exceeded) and a series of simple feasible moves which produce a feasible solution.…”

“…MNS 0 -TS uses the classic ejection chains proposed in Sevaux et al (2014), i.e., ejection chains are not extended to other time intervals. Table 7 displays the solution value reached by GRASP, MNS 0 -TS and MNS-TS for our set of instances.…”

“…We have addressed the dynamic memory allocation problem with a greedy randomized adaptive search (GRASP) in Sevaux et al (2014) and applied iterative approaches in .…”