System Bits: Sept. 17

Multicore memory management
According to MIT researchers, it may be time to let software rather than hardware manage high-speed on-chip memory caches.

Traditionally, managing the caches has required fairly simple algorithms that can be hard-wired into the chips but as multiple cores in SoCs proliferate, cache management becomes much more difficult.

As such, MIT’s Department of Electrical Engineering and Computer Science believes it is time to turn cache management over to software, and recently presented a new system, dubbed Jigsaw, that monitors the computations being performed by a multicore chip and manages cache memory accordingly.

In experiments simulating the execution of hundreds of applications on 16- and 64-core chips, the MIT researchers found that Jigsaw could speed up execution by an average of 18% — with more than twofold improvements in some cases — while actually reducing energy consumption by as much as 72%. They believe the performance improvements offered by Jigsaw should only increase as the number of cores does.

In most multicore chips, each core has several small, private caches. But there’s also what’s known as a last-level cache, which is shared by all the cores and is on the order of 40 to 60% of the chip. It is a significant fraction of the area because it’s so crucial to performance. If that cache were not there, some applications would be an order of magnitude slower, the researchers noted. Physically, the last-level cache is broken into separate memory banks and distributed across the chip. For any given core, accessing the nearest bank takes less time and consumes less energy than accessing those farther away but because the last-level cache is shared by all the cores, most chips assign data to the banks randomly.

Jigsaw, by contrast, monitors which cores are accessing which data most frequently and, on the fly, calculates the most efficient assignment of data to cache banks. For instance, data being used exclusively by a single core is stored near that core, whereas data that all the cores are accessing with equal frequency is stored near the center of the chip, minimizing the average distance it has to travel. Jigsaw also varies the amount of cache space allocated to each type of data, depending on how it’s accessed. Data that is reused frequently receives more space than data that is accessed infrequently or only once.

In principle, optimizing cache space allocations requires evaluating how the chip as a whole will perform given every possible allocation of cache space to all the computations being performed on all the cores. That calculation would be prohibitively time-consuming, but by ignoring some particularly convoluted scenarios that are extremely unlikely to arise in practice, the researchers were able to develop what they said is an approximate optimization algorithm that runs efficiently even as the number of cores and the different types of data increases dramatically.

Ultraviolet LED for portable, low-cost devices
With commercial uses for ultraviolet (UV) light on the rise, a new kind of LED under development at The Ohio State University could lead to more portable and low-cost uses of the technology.

Researchers said the patent-pending LED creates a more precise wavelength of UV light than today’s commercially available UV LEDs, and runs at much lower voltages and is more compact than other experimental methods for creating precise wavelength UV light. This could lend itself
to applications for chemical detection, disinfection, and UV curing. With significant further development, it might someday be able to provide a source for UV lasers for eye surgery and computer chip manufacturing.

The Ohio State engineers created LEDs out of semiconductor nanowires which were doped with the rare earth element gadolinium. The unique design enabled the engineers to excite the rare earth metal by passing electricity through the nanowires although they didn’t set out to make a UV LED. As far as they knew, nobody had ever driven electrons through gadolinium inside an LED before. They just wanted to see what would happen.

When the researchers started creating gadolinium-containing LEDs in the lab, they utilized another patent-pending technology they had helped develop—one for creating nanowire LEDs. On a silicon wafer, they tailored the wires’ composition to tune the polarization of the wires and the wavelength, or color, of the light emitted by the LED. Gadolinium was chosen not to make a good UV LED, but to carry out a simple experiment probing the basic properties of a new material they were studying, called gadolinium nitride. During the course of that original experiment, they noticed that sharp emission lines characteristic of the element gadolinium could be controlled with electric current.

Different elements fluoresce at different wavelengths when they are excited, and gadolinium fluoresces most strongly at a very precise wavelength in the UV, outside of the range of human vision. The engineers found that the gadolinium-doped wires glowed brightly at several specific UV frequencies.

Exciting different materials to generate light is nothing new, but materials that glow in UV are harder to excite. The only other reported device which can electrically control gadolinium light emission requires more than 250 volts to operate. The Ohio State team showed that in a nanowire LED structure, the same effect can occur, but at far lower operating voltages: around 10 volts. High voltage devices are difficult to miniaturize, making the nanowire LEDs attractive for portable applications.

The other device needs high voltage because it pushes electrons through a vacuum and accelerates them, just like a cathode ray tube in an old-style TV. The high-energy electrons then slam into gadolinium atoms, which absorb the energy and re-emit it as light in the UV. They believe their device works at significantly less voltage precisely because of the LED structure, where the gadolinium is placed in the center of the LED, exactly where electrons are losing their energy. The gadolinium atoms get excited and emit the same UV light, but the device only requires around 10 volts and because the LED emits light at specific wavelengths, it could be useful for research spectroscopy applications that require a reference wavelength, and because it requires only 10 volts, it might be useful in portable devices.

~Ann Steffora Mutschler