Dynamic decentralized mapping of tree-structured applications on NoC architectures

Abstract

This paper presents a novel application-driven and resourceaware mapping methodology for tree-structured streaming applications onto NoCs. This includes strategies for mapping the source of streaming applications (seed point selection), as well as embedding strategies so that each process autonomously embeds its own succeeding tasks. The proposed embedding strategies only consider the local view of neighboring cells on the NoC which allows to significantly reduce computation and monitoring overhead. Our vision is that this approach facilitates self-organizing embedded systems that provide the flexibility and fault-tolerance required in future silicon technologies. The results provided in this paper show that our local and decentralized algorithms can compete with previously presented global and centralized algorithms.

Smart hill climbing for agile dynamic mapping in many-core systems

“…Consequently, as can be seen, the communication paths between dispersed nodes will be stretched and can be congested by communications of other existing applications (1 to 3) . As a decentralized mapping algorithm for tree-structured applications, Weichslgartner et al [8] suggest three different methods for the first node selection. Two of them need prior knowledge about all incoming applications, while the one which is more dynamic, the farthest-away algorithm, leads to inefficient results as it does not consider the size of applications.…”

“…if openDir(ncur) = (0,0) then (7) moveDir ← a random direction; (8) else (9) moveDir ← (SF(ncur) < |T| ) ? openDir (ncur): -openDir(ncur); (10) divert moveDir by (yawAngle); (11) end if (12) nnext ← move in moveDir; (13) if nnext is available AND preferred to ncur then (14) ncur ← nnext ; (15) yawAngle ← 0 ; (16) else (17) yawAngle ← a random in opposite side ; (18) end if (19) end while (20) if ncur is preferred to nfn then (21) nfn ← ncur ; function.…”

Comparative analysis of dynamic task mapping heuristics in heterogeneous NoC-based MPSoCs

“…Global agents (GAs) store information about all the clusters of the NoC and use a negotiating policy with CAs in order to define to which cluster an application will be mapped. Another distributed approach is proposed in [4], which explores different implementations of a decentralized self-embedding algorithm, aiming to minimize network contention and latency while providing fault-tolerance support for NoC-based systems.…”

“…load of new applications) and/or performance (e.g. change the frequency operation for optimizing battery lifetime) requirements, which leads in both time-changing processor workload and communication patterns [1][3] [4]. Thus, offlinemapping techniques can be sub-optimal or inadequate in many scenarios.…”

Reconfigurable computing for network function virtualization: A protocol independent switch

“…Instead of locating all functions in one reconfigurable region, the research in [13] propose to interconnect reconfigurable regions with less resources to deliver a chain of functions. Network on chip (NoC) is a common interconnect architecture to establish communication between these reconfigurable regions [14].…”