Multicore-based Task Scheduling

People's demand for the computer's computing is always growing. It increased computer's computing the main method is to improve single processor's performance and using many processor.Improving single processor's performance there are two main ways; first, raises the basic frequency; second, enhances Instruction Per Cycle (IPC). Due to restrictions on power and frequency, Further raises the single processor basic frequency is limited. Enhances the IPC mainly have the time interleaving as Pipelines, the resource replication as Superscaler, these are Instruction Level Parallel (ILP), because the structure dependence, data dependence, control dependence, Further Enhances IPC is limited.In recent years, The researchers and the processor manufacturer will enhance computer's computing the method to shift to using many processor, have adopted parallel architecture as multicore, multiprocessor, cluster and so on.This paper first Summarized enhances the single processor performance method and analyzed the causes of these methods is limited. Classed the many processor's architecture by coupling and memory access model, analyzed the many processor structure meets subject matter:In operating system's many processor's task scheduling, support parallel programming language, support parallel programming technique, support parallel compiler and optimization techniques. Many processor's task scheduling is main research question in this paper.This paper started from the single processor's task scheduling, Compared the single processor and many processor task scheduling differences. Many processor's task scheduling's key question is the load balancing, Load balancing difficulty is load criteria choice, migrated task choice and so on. many processor's task scheduling method divides according to the task allocation opportunity into the static task scheduling and the dynamic task scheduling.With the wide application of multicore processors, this paper as multicore for task scheduling environment model dissect the multicore task scheduling, summed up the domestic and international research in this field, analyzed the static and dynamic task scheduling model and scheduling algorithm.The paper based on open source Linux kernel 2.6 scheduling algorithm, Detailed the Linux kernel 2.6 task scheduling algorithm and related data structure, and Linux kernel 2.6 support of multicore. Pointed out the Linux kernel 2.6 needs to change in the multicore task scheduling, Proposed improved algorithm. According to this algorithm changed Linux kernel 2.6 task scheduling. The comparison test results show that the improved algorithm enhanced the efficiency of multicore task scheduling, raised the system performance.