Font Size: a A A

The Research And Design Of Load Balancing Method In Multi-core Environment

Posted on:2012-08-15Degree:MasterType:Thesis
Country:ChinaCandidate:J Z LiFull Text:PDF
GTID:2178330332492371Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the development of multi-core computers, parallel computing has become the primary choice of high performance computing. Many large-scale computing tasks can be completed by multi-core computers. But now most of the programming work still use the traditional serial programming method, which causes the process can not make good use of the excellent hardware performance of the multi-core systems. Particularly in the high precision reconstruction of complex surface, the speed of the traditional serial process has become a bottleneck in the practical application. It is not difficult to find out the load balancing research of high precision reconstruction in multi-core environment has an important meaning.There are two parts which consume lots of system time in the high precision reconstruction of complex surface system. One is reading mass data, the other is high precision reconstruction. So how to improve the performance of the two parts is the key to break the speed bottlenecks of the system.First, using the quick-read means of mass data based on multi-core environment to read data from memory and secondary memory. The method uses static load balancing, dynamic load balancing and a division method based on the size of view map to achieve high-speed parallel extraction of mass data in multi-core environment.Second, choosing Bezier surface which is an important surface in computer graphics and researching fast-generate algorithm of Bezier surface based on multi-core environment. This method designs the load balancing module of high-order and high-density Bezier surface and parallel the generation of Bezier surface by the strategy of matrix storage structure based on the priority of cache. This algorithm programs with OpenMP and achieves higher efficiency.At last, experiments show that the quick-read means of mass data based on multi-core environment makes full use of the characteristics of multi-core environment and has good adaptability in the case of increasing the amount of data. Fast-generate algorithm of Bezier surface based on multi-core environment was tested to generate high degree and high density Bezier surface on the platform of dual-core processor,quad processor and eight-core processor.The tests show higher speedup and higher CPU utilization, which proves the good parallel efficiency and expansibility of the algorithm.
Keywords/Search Tags:Multi-core, Load Balancing, Speedup, Mass Data, Bézier Surface
PDF Full Text Request
Related items