Resource allocation method for task number and performance-aware reconfigurable multi-core processors

Abstract

A resource allocation method for reconfigurable multi-core processors with task number and performance awareness. Dynamic reconfigurable multi-core processors have the ability to dynamically reconfigure on-chip resources during runtime. In order to reduce the average turnaround time of tasks, improve system throughput and chip Resource utilization offers huge room for optimization. In the present invention, within each operating system scheduling interval, the resource allocator first allocates logic cores on average according to the number of tasks, and after running for a certain clock cycle, sorts them according to the performance of the tasks (reflecting the resource requirements of the tasks), and finds Assign tasks with small resource requirements, reduce the granularity of logical cores occupied, and allocate idle physical cores obtained from tasks with small resource requirements to tasks with high resource requirements, so as to increase the number of tasks with high resource requirements The granularity of logical cores occupied. When the current load of the system changes or the task itself enters a new running stage, the resource allocator will make timely adjustments in the next operating system scheduling to make full use of chip resources.

Description

technical field [0001] The invention relates to operating system design and runtime support in the field of computer system structure, in particular to a resource allocation method for reconfigurable multi-core processors with task number and performance awareness Background technique [0002] Introduction of dynamic reconfigurable multi-core processor technology and resource allocation algorithm based on dynamic reconfigurable multi-core processor [0003] Dynamic Reconfigurable Chip Multiprocessor (DRCMP) is a new type of multi-core processor architecture proposed by academic circles in recent years. The DRCMP chip is composed of a group of isomorphic physical processing units with simple structure, which we call physical cores. DRCMP can dynamically reconstruct a logical logical core composed of one or more physical cores at runtime. Tasks run on logical cores. The more physical cores that constitute a logical core, the greater the granularity of this logical core. Th...

AI Technical Summary

Problems solved by technology

If the efficiency of the task is greater than the maximum efficiency, it means that the logical core occupied by the task is too small to meet the demand, and the logical core is appropriately increased to reduce the efficiency

Images