Description method of reconfigurable hardware resource in scheduling

Abstract

The invention relates to a description method of reconfigurable hardware resource in scheduling, which is characterized in that on the basis of a reconfigurable computation model which is built based on a one-dimension part, a resource axis is adopted as a transverse axis, a time axis is adopted as a longitudinal axis, so an R-T coordinate system is established, a reconfigurable hardware task is described as a Ti formation task area, then boundary of a task under execution in the R-T coordinate system is identified, at the beginning, three boundaries exist in the R-T coordinate system, after one task obtains the resource, a rectangular area in the R-T coordinate system can be occupied so as to introduce a new boundary, the new boundaries restrict the arrangement of other task to be scheduled, when arranging the task, at least one of a bottom edge and a left edge or a right edge thereof is close to the existing boundary, then a correspondent boundary node diagram is generated through the mapping of a task boundary diagram, a boundary table data structural diagram is used for describing the node diagram of the boundary, and finally the task scheduling is realized. The description method has the advantages that the information record is less, the maintenance program is simple, and the searching time is short when in scheduling.

Description

technical field [0001] The invention belongs to the technical field of computers, and in particular relates to a method for describing reconfigurable hardware resources during scheduling. Background technique [0002] At present, the description methods for reconfigurable hardware resources during scheduling mainly include the free area description method and the description method of the state matrix. The free area description method mainly records the free area to describe the resources, and the most important one is the MER maximum free rectangle method. (MaximalEmpty Rectangle), the MER method records a non-overlapping free resource rectangle list, and obtains the placement location of the hardware task by finding the matching free resource rectangle. This method needs to maintain the resource rectangle list, and it is complicated to find the matching free resource High degree; the state matrix description method manages resources by maintaining a matrix that records the...

AI Technical Summary

Problems solved by technology

[0002] At present, the description methods for reconfigurable hardware resources during scheduling mainly include the free area description method and the description method of the state matrix. The free area description method mainly records the free area to describe the resources, and the most important one is the MER maximum free rectangle method. (MaximalEmpty Rectangle), the MER method records a non-overlapping free resource rectangle list, and obtains the placement location of the hardware task by finding the matching free resource rectangle. This method needs to maintain the resource rectangle list, and it is complicated to find the matching free resource High degree; the state matrix description method manages resources by maintaining a matrix that records the state of the reconfigurable computing unit (Reconfigurable Computing Unit), and scans the state matrix to find the place where the hardware task can be placed. This method needs to scan and find the state matrix, which costs It takes a long time, and these two methods are suitable for running in the reconfigurable computing model based on two-dimensional devices, because in the two-dimensional device model, computing resources are divided and organized in a two-dimensional array, and the minimum unit of reconstruction is A unit in a two-dimensional array, a hardware task occupies a rectangular area in the array, task placement has a two-dimensional degree of freedom, if the task time attribute is considered when scheduling hardware tasks, the running time of the two-dimensional device model will be longer , the maintenance cost will be higher, which is not conducive to the real-time task scheduling of the system

Images