A Scheduling Algorithm For Heterogeneous Multi-core Processors And Its Implementation

A heterogeneous multi-core architecture is a chip multiprocessor composed ofcores of varying size, performance, and complexity. This architecture can providesignificantly highter performance in the same area than a conventional chipmultiprocessor (homogeneous). For efficient utilization of these performane asymmetricmulti-core processors, application threads must be assigned to cores such that theresource needs of a thread closely matches resource availability at the assigned core.Through analysis of existing scheduling algorithms for heterogeneous, this thesisproposes a new task scheduling algorithm on a heterogeneous architecture. It comparesa thread’s instruction numbers in the two continuous stage with the same time in thesame core, to determine the conditions of thread migration, then it will finish schedulingby assigning the cpu-bound thread to the fast core, and the memory-bound thread to theslow core.Then the scheduling algorithm and its implementation in Linux2.6kernel arestudied in detail. And our algorithm is implemented by modifying the Linux2.6.21kenel.At last, I compare the performance between the modifyed kernel with the stock kernelby using SPEC2006(Standard Performance Evaluation Corporation). The experimentindicates that our algorithm can get a performance improvement of5.14%at a goodsituation for different benchmarks.