Optimizing the Performance of Wintel Applications

The best medicine for most sorts of performance problems is invariably preventative. Despite advance in software performance engineering [1,2], developing complex computer programs that are both functionally correct and efficient remains a difficult and time-consuming task. This paper specifically looks at tuning Windows NT applications running on Intel hardware from the perspective of optimizing processor cycles and resource usage. Fine-tuning the execution path of code remains one of the fundamental disciplines of performance engineering. To bring this topic into focus, I will describe a case study where an application designed and developed specifically for the Microsoft Windows NT environment is subjected to a rigorous analysis of its performance using several commercially available CPU execution profiling tools. Since one of the development tools used to optimize the application program under consideration requires an understanding of the internal workings of Intel processors, this will justify an excursion into the area of Intel processor hardware performance. An application tuning case study. The application that is the target of this analysis is a C language program that was written to collect Windows NT performance data continuously on an interval basis. The performance of this application is quite important. Since the app is designed primarily for use as a performance tool, it is very important that it run efficiently. A tool designed to diagnose performance problems should not itself be responsible for causing performance problems. Moreover, the customers for this application, many of whom are experienced Windows NT performance analysts, are a very demanding group of users. Figure 1 illustrates the application structure and flow, which is straightforward. Following initialization, the program we developed enters a continuous data collection loop. Inside this loop, Windows NT services are called to retrieve selected performance data across a well-documented Win32 interface. The program, which consisted of a single executable module called HQTIVJWWI\I, simply gathers performance statistics across this interface and logs the information collected to a local disk file. It is the optimization of the code within this inner loop that was the focus of this study.