| Regarding the multi-core processors are mroe and more popular, most application developers will face with great opportunities and challenges. In the past, most developers usually can only develop procedure on single core platform in the lack of parallel development environment. In most case, it would not consider the concurrent execution of program. However, with the time passing, it can foresee that the single-core processor will be outdated. In recent years, multi-core processor has great developted and popularized, most developers own multi-core development enviroment. If not pay more attention on program concurrent execution in developing, it will not play the concurrent execution feature of multi-core platform, even further improve program efficiecy.In addition, along with the Internet speed up development, search engine become the most popular application in Internet. Search engine is a common application of full-text retrieval system. As multi-core platform popularization and hardware price reducing, it wil be possible to creat a suitable to industry characteristics retrieval system. The research for full-text Index model, mainly through the modification of index structure to improve the performance of the index. In this document, it will present another way to improve full-text retrieval system dynamic successive tree index model algorithm which based on multi-core platform parallelization, and then, improve the efficiency of index generation and retrieval. The document works on the following:1.Study the main features of multi-core processor and questions of multi-core programming, compare the differents between muti-core platform parallel computing and traditional multi-machine distributed parallel computing, which will be a guidance for the research of parallel algorithm.2.Study the main features and differents index type of full-text retrieval system, compare with subsequent dynamic index tree. According to the generating speed, retrieval efficiency and upgrading complexity, choose the best Index type dynamic subsequent tree index algorithm to do the research.3.Consider the feature of multi-core programing to design a task allocation way of Dynamic successive tree index generation algorithm. Each parallel thread can work with the most reasonable way to play the best performance of program, and add OpenMP guidance statement in Dynamic following original tree generation algorithm and Participle retrieval algorithm to achieve parallel program.4. Compare the laboratory result with Dynamic following original tree generation algorithm, modified Parallel generating algorithm, Dynamic tree points following word to find the original algorithm and More than words in parallel search algorithm, which prove the operating efficiency of parallel processed algorithm get significant improvement on multi-core plarform.5. Successive Trees original dynamic parallel algorithm modified to create algorithms, and dynamic successor tree multisection word Find the original algorithm with multisection word parallel search algorithm to compare the experimental results prove that after the parallel processing algorithms on multi-core platformsignificantly improve operating efficiency. |
PDF Full Text Request