Ixa Architecture-based Network Programming Parallel Optimization Study

With the booming expanding of the Internet, requests are increasing rapidly for high reliability and good adaptation of network equipments. As a newly-emerged special processor, network processor comes in being in time, which generally utilizes specialized instruction set, multi-core and multi-threading to get optimized processing performance of packet, though in the same time the complexity of its architecture brings big challenges for programming.A kind of parallelizing compiler developed by Intel based on IXA is introduced in detail for its design and performance analysis to make researching step on the performance optimization for parallelizing execution of network applications.Investigation and analyzing has been made by the science and the industry for architecture of the network processor and network applications to testify the high benefits that could be brought by good combination of the logic of network application and the parallel infrastructure of the network processor, which confirms that the logical parallelization in network applications and the paralleled hardware structure of network processor are the promising basics for potential excavating of network hardware and developing of high quality network applications.IXP C, as a parallelizing compiler, is designed aiming at parallel maximization in the sequentially semantic applications by parallelizing them and mapping them to the hardware units of the network processor, which abstract away most of the hardware details for programmers to let them focusing on logics of the programs, facilitates the development of high performance network applications executing on network processors, and thus brings good portability and expandability for the programs. Multiprocessing/Multithreading(MTP), Context Pipelining(CTX) and Pasting are three of the main mapping forms of IXP C complier. Among these, the context pipelining algorithm in CTX is the most important technique for maximizing the parallelization, which will be explicated deeply for its principle and performance and also the performance comparisons of MTP and CTX are analyzed.Some shortcomings are observed in the context pipelining algorithm of IXP C...