Brute force almost always beats elegance when it comes to processor power. Increase the clock speed, super-cool the cpu, add more cores, and the like have been the major processing power increase techniques for some time. However, researchers from the North Carolina State University are taking the elegance route on multicore CPUs. Their focus is on prefetching data. Processors predict what data may be needed next, and load it into its cache. In some cases, it may load the wrong data and slow down performance as it seeks the correct set of data. In other words, prefetching could hinder performance. Dr. Tan Solihin, associate professor of electrical and computer engineering, has a solution, better prediction and the ability to turn off prefetching when it may actually be cumbersome.
He explains, "The first technique relies on criteria we developed to determine how much bandwidth should be allotted to each core on a chip. By better distributing the bandwidth to the appropriate cores, the criteria are able to maximize system performance. The second technique relies on a set of criteria we developed for determining when prefetching will boost performance and should be utilized, as well as when prefetching would slow things down and should be avoided."
The results were shown to increase performance by 40% on processors without prefetching, as expected, and 10% in the processors that utilize prefetching at all times. Read more about his system in the paper, "Studying the Impact of Hardware Prefetching and Bandwidth Partitioning in Chip-Multiprocessors."
Eavesdropper