Abstract:

High performance computing is widely used in science and engineering to solve large scale computation problems. But the sustained performances achieved for the real applications do not increase as fast as the peak performances do. In fact， the sustained performance is a only about 5~10% of the peak performance, and the gap between the sustained performance and the peak performance is widening. Code performance optimization, which is one of the effective ways to solve this problem, draws the attentions of the research community. Based on Itanium microprocessors, this paper summarizes the general methods for code performance optimization and gives the common steps for code performance optimization and analysis. According to the steps, the performances for four codes are analysed in detail to find the performance bottlenecks and the key subroutine codes. Then four codes are optimized in the Itanium microprocesspor, using the code optimization techniques based on cache and instruction pipeline. Finally, the test results for the four performance optimization codes show that the performances are increased by 8~33% respectively.

Key words: itaniuam microprocessor;code performance optimization;code performance analysis