From April to June, 2012, School of Computer Science and Technology (SCST) opened the advanced course on computer architecture---“Software Performance Optimization for Multi-core Architectures”. Dr. TANG Xinan, senior engineer of Intel Compiler Lab (U.S.) and guest professor of SCST, was invited as the lecturer. This course gained supports from USTC Graduate Student Education Innovation Plan Project and Intel Corporation.
This course was mainly opened to the graduate and senior undergraduate students in SCST, and all the students who were interested in software performance optimizations. The course covered related knowledge on professional courses including Computer Architecture, Compiler Principles, Operation System, and Computer Networks, etc., focusing on how to utilize existing knowledge to optimize software performance for Multi-core architectures.
The course firstly introduced the main factors that may deteriorate program performance, and then explained in detail the basic optimization strategies and methods, and the advanced compiler optimization techniques such as loop unrolling and data pre-fetching. After that, the course gave an introduction to the Instruction Set Architecture(ISA), thoroughly described Instruction Pipelining and Instruction Scheduling, and then further explained advanced compiler optimization technologies. Furthermore, after the study on memory hierarchy and CPU cache, the course put the emphasis on Cache Coherence and Cache Consistency in the Multi-core Architecture, plus effective solutions for common problems.
The course was mainly offered by classroom teaching, while extra learning and practice tasks were also arranged for students. Combined with the curricular and extra study, the students could deeply understand the necessity of software performance optimization and master lots of key software optimization technologies. They expressed that the course had broadened their vision, enhanced their professional skills, and improved their abilities in solving problems.
This course contained 8 lectures and the whole lecturing process was filmed. The video, courseware and related learning materials can be downloaded on the teaching homepage: http://home.ustc.tsg211.com/~shengjie. The video and courseware can also be viewed on the web site: http://202.38.95.18/view.asp?categoryid=2&movieid=2001 .
