Engineers Increase Computer Program Speed, Retain Safety

(Pacific Northwest National Laboratory)

Researchers at North Carolina State University in Raleigh have developed a software tool that helps computer applications run more efficiently without sacrificing safety features that previously slowed down their performance. The faculty/student team will discuss this process on 5 April at the International Symposium on Code Generation and Optimization in Chamonix, France.

Application software programs are adding more safety features to protect users, but those features can also slow the programs down by 1,000 percent or more. Because of this overhead, software developers are tempted to leave out the safety features. NC State engineering professor James Tuck says, however, “Leaving out those features can mean that you don’t identify a problem as soon as you could or should,” adding “particularly if it’s a problem that puts your system at risk from attack.”

These safety features have in the past been put directly into a software program’s code, and run through the same central processing unit (CPU) chip that the program itself uses. This sharing of the CPU of main program code and safety modules is what slows the program down.

Many of the newer computer systems, however, have multi-processing chips that divide the CPU into four to eight units or cores. The NC State engineers developed a tool that takes advantage of multi-core computer chips by running the safety features on a separate core in the same chip, which allows the main program to run at close-to-normal speed.

Other multi-core approaches have been tried before, but involved replicating huge chunks of code among the cores, a time-consuming process that used a great deal of power. Tuck says the new parallelization tool streamlines the safety feature work done by other cores, and requires no manual reprogramming.

Tuck and Ph.D. student Sanghoon Lee implemented the tool as a plug-in for the Gnu Compiler Collection of open-source software, and extending the tool to support a wider range of applications and functions. The research was supported by the National Science Foundation.

*     *     *