A Mapping Decision Method for Reconfigurable Architecture Based on Cost Computation

Abstract

The invention discloses a reconfigurable architecture mapping decision-making method based on expense calculation. The method includes: building four expense models which are delay expense, interconnection expense, PE occupancy rate expense and similarity expense according to data dependence graph (DDG) indicating application algorithm core circulation and a reconfigurable architecture; for multiple feasible mapping schemes existing in one operating node, calculating the corresponding expense value, in the four expense models, of each feasible mapping scheme; sequentially traversing each feasible mapping scheme from primary to secondary according to the degree of influence of the four expense models on the mapping results, and gradually diminishing the feasible mapping scheme set to obtain the optimal mapping scheme. By the decision-making method, the factors with large influence on mapping are more predominated in mapping decision making. By the expense models and the mapping decision-making method, configuration information with high efficiency can be obtained, and parallelism of the reconfigurable architecture can be brought into full play.

Description

technical field [0001] The invention belongs to the field of embedded information technology, in particular to a reconfigurable architecture-oriented mapping decision-making method based on cost calculation. Background technique [0002] The array configuration information used by the reconfigurable architecture not only ensures the correct function of the application program in the reconfigurable architecture, but also takes into account the limited hardware resources of the reconfigurable array for efficient execution. How to stably and effectively automatically generate configuration information that meets the requirements has always been an unsolvable problem. [0003] The problem of mapping the core loop in the application program to the reconfigurable array has been proved to be a non-deterministic (Non-deterministic Polynomial, NP for short) complete problem of polynomial complexity, by Johnson DS.1982 in "Journal of Algorithms "Proof of the paper "The NP-completenes...

AI Technical Summary

Problems solved by technology

[0005] Technical problem to be solved: Aiming at the deficiencies in the prior art, the present invention provides a new type of overhead model, mapping overhead function and mapping decision-making method suitable for reconfigurable architectures, for evaluating reconfigurable architecture software pipeline overhead and Determine the best mapping scheme, and solve the technical problems in the prior art that are difficult to establish an accurate overhead model and determine the best mapping

Images