The Kynix Blog

In today's computer chips, memory management is based on what computer scientists call the principle of locality: If a program needs a chunk of data stored at some memory location, it probably needs the neighboring chunks as well.But that assumption breaks down in the age of big data, now that computer programs more frequently act on just a few data items scattered arbitrarily across huge data sets. Since fetching data from their main memory banks is the major performance bottleneck in today's chips, having to fetch it more frequently can dramatically slow program execution.This week, at the International Conference on Parallel Architectures and Compilation Techniques, researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) are presenting a new programming language, called Milk, that lets application developers manage memory more efficiently in programs that deal with scattered data points in large data sets.In tests on several common algorithms, programs written in the new language were four times as fast as those written in existing languages. But the researchers believe that further work will yield even larger gains.The reason that today's big data sets pose problems for existing memory management techniques, explains Saman Amarasinghe, a professor of electrical engineering and computer science, is not so much that they are large as that they are what computer scientists call "sparse." That is, with big data, the scale of the solution does not necessarily increase proportionally with the scale of the problem."In social settings, we used to look at smaller problems," Amarasinghe says. "If you look at the people in this [CSAIL] building, we're all connected. But if you look at the planet scale, I don't scale my number of friends. The planet has billions of people, but I still have only hundreds of friends. Suddenly you have a very sparse problem."Similarly, Amarasinghe says, an online bookseller with, say, 1,000 customers might like to provide its visitors with a list of its 20 most popular books. It doesn't follow, however, that an online bookseller with a million customers would want to provide its visitors with a list of its 20,000 most popular books.Thinking locallyToday's computer chips are not optimized for sparse data—in fact, the reverse is true. Because fetching data from the chip's main memory bank is slow, every core, or processor, in a modern chip has its own "cache," a relatively small, local, high-speed memory bank. Rather than fetching a single data item at a time from main memory, a core will fetch an entire block of data. And that block is selected according to the principle of locality.It's easy to see how the principle of locality works with, say, image processing. If the purpose of a program is to apply a visual filter to an image, and it works on one block of the image at a time, then when a core requests a block, it should receive all the adjacent blocks its cache can hold, so that it can grind away on block after block without fetching any more data.But that approach doesn't work if the algorithm is interested in only 20 books out of the 2 million in an online retailer's database. If it requests the data associated with one book, it's likely that the data associated with the 100 adjacent books will be irrelevant.Going to main memory for a single data item at a time is woefully inefficient. "It's as if, every time you want a spoonful of cereal, you open the fridge, open the milk carton, pour a spoonful of milk, close the carton, and put it back in the fridge," says Vladimir Kiriansky, a PhD student in electrical engineering and computer science and first author on the new paper. He's joined by Amarasinghe and Yunming Zhang, also a PhD student in electrical engineering and computer science.Batch processingMilk simply adds a few commands to OpenMP, an extension of languages such as C and Fortran that makes it easier to write code for multicore processors. With Milk, a programmer inserts a couple additional lines of code around any instruction that iterates through a large data collection looking for a comparatively small number of items. Milk's compiler—the program that converts high-level code into low-level instructions—then figures out how to manage memory accordingly.With a Milk program, when a core discovers that it needs a piece of data, it doesn't request it—and a cacheful of adjacent data—from main memory. Instead, it adds the data item's address to a list of locally stored addresses. When the list is long enough, all the chip's cores pool their lists, group together those addresses that are near each other, and redistribute them to the cores. That way, each core requests only data items that it knows it needs and that can be retrieved efficiently.That's the high-level description, but the details get more complicated. In fact, most modern computer chips have several different levels of caches, each one larger but also slightly less efficient than the last. The Milk compiler has to keep track of not only a list of memory addresses but also the data stored at those addresses, and it regularly shuffles both around between cache levels. It also has to decide which addresses should be retained because they might be accessed again, and which to discard. Improving the algorithm that choreographs this intricate data ballet is where the researchers see hope for further performance gains."Many important applications today are data-intensive, but unfortunately, the growing gap in performance between memory and CPU means they do not fully utilize current hardware," says Matei Zaharia, an assistant professor of computer science at Stanford University. "Milk helps to address this gap by optimizing memory access in common programming constructs. The work combines detailed knowledge about the design of memory controllers with knowledge about compilers to implement good optimizations for current hardware."  

Imec, the Belgian nanoelectronics research center, will present at this week's 'CMOS Image Sensors for High Performance Applications' workshop in Toulouse (France) a prototype of a high-performance, time-delay-integration (TDI) image sensor. The image sensor is based on imec's proprietary embedded charge-coupled device (CCD) in CMOS technology. Imec developed and fabricated the sensor for the French Space Agency, CNES, which plans to utilize the technology for space-based earth observation.  The prototype image sensor combines a light-sensitive, CCD-based TDI pixel array with peripheral CMOS readout electronics. By integrating CCD with CMOS technology, imec combined the best of both worlds. The CCD pixel structure delivers low-noise TDI performance in the charge domain, while CMOS technology enables low-power, on-chip integration of fast and complex circuitry readouts.A TDI imager is a linear device that utilizes a clever synchronization of the linear motion of the scene with multiple samplings of the same image, thereby increasing the signal to noise ratio. CCDs fit extremely well with the TDI application since they operate in the charge domain, enabling the movement of charges without creating excess noise. By combining the TDI pixels array with CMOS readout circuitry on the same die, imec produced a camera-on-a-chip or system-on-a-chip (SOC) imager, which reduces the overall system complexity and cost. The CMOS technology enables on-chip readout electronics, such as clock drivers and analog-to-digital convertors (ADCs), operating at higher speeds and lower power consumption not possible with traditional CCD technology.The prototypes were fabricated using imec's 130nm process with an additional CCD process module. An excellent charge transfer efficiency of 99.9987 % has been measured ensuring almost lossless transport of charges in the TDI array, and guaranteeing high image quality. Imec's specialty imaging platform combines custom design (i.e., specialized pixels, high-performance readout circuits and chip architectures) with optimized silicon processing, such as dedicated implants and backside thinning, to achieve high-end specialized imagers.  

At SPIE Photonics West, imec will present a new set of snapshot hyperspectral CMOS image sensors featuring spectral filter structures in a mosaic layout, processed per-pixel on 4x4 and 5x5 'Bayer-like' arrays.Imec's hyperspectral filter structures are processed at wafer-level on commercially available CMOS image sensor wafers, enabling extremely compact, low cost and mass-producible hyperspectral imaging solutions. This paves the way to multiple applications ranging from machine vision, medical imaging, precision agriculture to higher volume industries such as security, automotive and consumer electronic devices."Imec's latest achievements in hyperspectral imaging emphasize how our promising technology has become an industrially viable solution for a number of applications," said Andy Lambrechts, program manager at imec. "The new mosaic architecture, and extended spectral range, brings unique advantages compared to our previously announced hyperspectral linescan sensors for applications in which scanning would not be practical. It enables spectral imaging in a truly compact, tiny form-factor, that can even be scaled to handheld devices. From the technology standpoint, we have now successfully demonstrated linescan and tiled sensors, in which spectral filters cover many pixels, to mosaic sensors, in which filters vary from pixel to pixel. At the same time, the spectral range is extended and now covers down to 470nm."The newly developed mosaic sensors feature one spectral filter per pixel, arranged in mosaics of 4x4 (16 spectral bands) or 5x5 (25 spectral bands) deposited onto a full array of 2 Million pixels 5.5µm size CMOSIS CMV2000 sensor. Two versions of the mosaic hyperspectral image sensors have been developed:one 4x4 mosaic with 16 bands in the 470-630nm (visible range)one 5x5 mosaic with 25 bands in the 600-1000nm range (Visible – NIR range)"Imec's hyperspectral imaging sensors (100bands linescan, 32bands tiled and 16/25bands mosaic designs) are off-the-shelf, commercially available engineering sample sensors that we developed to address the fragmented machine vision market and to trigger interest for this unique technology from potential end-users in other industries," explained Jerome Baron, business development manager at imec. "We also offer customized spectral filtering solutions for companies that are already familiar with the technology and interested in developing proprietary solutions with a specific performance in terms of speed, compactness, spatial versus spectral resolution, bands selection, or cost."Related products:ANPVC5030ANPVC2260ANPVC1470ANPVC1210 

University of Utah electrical engineers fabricated the smallest plasma transistors that can withstand high temperatures and ionizing radiation found in a nuclear reactor. Such transistors someday might enable smartphones that take and collect medical X-rays on a battlefield, and devices to measure air quality in real time.     "These plasma-based electronics can be used to control and guide robots to conduct tasks inside the nuclear reactor," says Massood Tabib-Azar, a professor of electrical and computer engineering. "Microplasma transistors in a circuit can also control nuclear reactors if something goes wrong, and also could work in the event of nuclear attack." A study of the new transistors by Tabib-Azar and electrical engineering doctoral student Pradeep Pai appears online Thursday, March 20 in the journal IEEE Electron Device Letters, published by the Institute of Electrical and Electronics Engineers. The study was funded by the Defense Advanced Research Projects Agency. Transistors are the workhorses of the electronics industry. They control how electricity flows in devices and act as a switch or gate for electronic signals. Billions of transistors are typically fabricated as individual but connected components on a single computer chip. The most commonly used type of transistor is called a metal oxide semiconductor field effect transistor, or MOSFET. Transistors control the flow of electrical charge through a silicon channel using an electric field to turn the transistor on or off, similar to a valve with the electric field as its control knob and electric charge as its current flow. Silicon-based transistors are a crucial component in modern electronics, but they fail above 550 degrees Fahrenheit – the temperature at which nuclear reactors typically operate. Plasma-based transistors, which use charged gases or plasma to conduct electricity at extremely high temperatures, are employed currently in light sources, medical instruments and certain displays under direct sunlight (but not plasma TVs, which are different). These microscale devices are about 500 microns long, or roughly the width of five human hairs. They operate at more than 300 volts, requiring special high-voltage sources. Standard electrical outlets in the United States operate at 110 volts. The new devices designed by the University of Utah engineers are the smallest microscale plasma transistors to date. They measure 1 micron to 6 microns in length, or as much as 500 times smaller than current state-of-the-art microplasma devices, and operate at one-sixth the voltage. They also can operate at temperatures up to 1,450 degrees Fahrenheit. Since nuclear radiation ionizes gases into plasma, this extreme environment makes it easier for plasma devices to operate. "Plasmas are great for extreme environments because they are based on gases such as helium, argon and neon that can withstand high temperatures," says Tabib-Azar. "This transistor has the potential to start a new class of electronic devices that are happy to work in a nuclear environment." A conventional transistor is made with two active layers, one on top of the other. Electricity flows through one of the layers, called the channel. The other layer, called the gate, controls current flowing in the channel. If sufficient voltage is applied to the gate, the transistor turns on. For the new study, Tabib-Azar and Pai deposited layers of a metal alloy to form the gate on a 4-inch glass wafer. A layer of silicon then was deposited on top of the gate. Unlike typical transistors, the Utah microplasma transistor "channel" is an air gap that conducts ions and electrons from the plasma once a voltage is applied. To achieve this unique design, the team etched away portions of the silicon film using a chemically reactive gas. This etching process leaves behind cavities and empty spaces to form the transistor's channel and expose the gate underneath. The channel tested in this new study was 2 microns wide and 10 microns long, and helium was used as the plasma source. "Although the length scales are much smaller here, we came up with an innovative way to make these structures three-dimensional," Tabib-Azar says. "We are currently connecting these devices to form logic gates and computing circuits that we will test in our experimental nuclear reactor at the University of Utah, a facility not found in most other universities." Traditional MOSFETs require metal to connect circuits, says Tabib-Azar, but the Utah microplasma devices will use a plasma-based connection to enable communication. As a result, these circuits will only be operational when powered up and will disappear otherwise, making them suitable for defense applications. These plasma devices could also be used as an X-ray imaging source in the next five years, says Tabib-Azar. Because the device dimensions are so small, X-ray images from a wounded soldier in the field could be collected on a smartphone equipped with transistors that also generate the X-rays, says Tabib-Azar. In another five years, the devices could be used to detect and identify aerosol pollutants based on the color emitted when the substance passes through the device. "These chemical sensing devices could be used to quantitatively monitor air quality in real time and enable researchers to construct an accurate air-quality map," he adds. In the nearer-term, these new transistors could be used to generate X-rays to draw fine lines in silicon to pattern microscale devices for the electronics industry. With this new X-ray technique, Tabib-Azar says, "you can do the same thing you would with laser printing, but instead you can use these tiny X-ray sources to print on a silicon wafer. This gives engineers the ability to do X-ray lithography without having to use very heavy lenses and X-ray beam shaping devices."  

With the ICL5101, Infineon Technologies AG extends its portfolio of lighting control ICs, addressing lighting systems in the range of 40W to 300W. The new high-voltage resonant controller IC provides a high level of integration which translates to a reduction in system cost. Typical applications which benefit from these features include indoor and outdoor LED lighting, high-bay and low-bay lighting, street lighting, parking garage and canopy lighting, office lighting, retail and shop lighting. Since the total cost of ownership is an important aspect for industrial lighting, customers prefer to use resonant topologies supported by the new ICL5101 due to its high efficiency up to 95%.  The highly integrated ICL5101 allows for advanced LED driver designs with approximately 25% less components compared to similar solutions which require separate PFC and resonant ICs. This leads to smaller form factors with more reliable designs, less complex PCB layouts and reduced costs. The ICL5101 integrates the half-bridge and the PFC gate drivers. All operation parameters of the IC are adjustable by simple resistors, enabling cost effective but reliable and stable parameter-settings.The chip supports outdoor use by an extended junction temperature ranging from -40°C to +125°C.The LED controller ICL5101 is designed to control resonant converter topologies such as LLC. The integrated digital PFC stage operates both in critical conduction mode (CrCM) and discontinuous conduction mode (DCM), which allows an extremely stable regulation in low load conditions, occurring for e.g. when the device is dimmed. The LED lighting can be dimmed down over an extremely wide range from 100% to 0.1% of its nominal load. State of the art dimming today typically ranges from 100% to 5%. In addition, the ICL5101 enables an ultra-fast time to light – under any conditions – with less than 200ms.The adjustable PFC stage of the ICL5101 delivers high power quality, providing a low total harmonic distortion (THD) of less than 10% and a high power factor of more than 0.99 over wide line input voltage range. This enables lighting manufacturers to comply with energy efficiency standards. Furthermore the output of the ICL5101 is extremely stable over line voltage variations. A comprehensive set of protection features including external over temperature protection and capacitive load protection ensure the detection of fault conditions and increase system safety.With the introduction of ICL5101 Infineon once again demonstrates its technology leadership for highly efficient driver solutions. Just recently, the ILD6150 step-down driver IC was nominated as finalists in the product category "ICs and electronic components" for the 2015 LEDs Magazine Sapphire Awards .  

What if you could stick an OLED panel on your wall with a magnetic mat? A detachable OLED (organic light-emitting diode) panel that would just as easily be taken off as stuck on the wall? Reports surfaced on Tuesday that South Korea-based LG Display has fashioned just the thing, a 0.97 mm thick 55" flat OLED TV panel and only 1.9 kilos (4.2 pounds). LG Display showcased the screen in Korea.By comparison, LG Display's existing 55-inch OLED panel is 4.3 mm thick. Engadget's Jon Fingas said that "it raises the possibility of big-screen sets that easily blend into your living room's décor." That's the good news. The sad news is that there is no word about when such displays will make it to retail shops.The unveiling was part of a broader announcement to showcase the company's plans for the future, which center on OLED tech, said Don Reisinger in CNET. The screen was presented as one of the company's future displays at the media event. Using a magnetic mat, the screen can easily be stuck to—or removed from— a wall. To remove the display from the wall, said Reisinger, you peel the screen off the mat. The Yonhap News Agency report carried a photograph of a model gently lifting the detachable wallpaper OLED panel at the event in Seoul on Tuesday. Yonhap News Agency referred to the LG Display screen as a "wallpaper OLED panel." Don Reisinger of CNET referred to it as press-on wallpaper TV.Strategically, the unveiling tells us something about LG Display, said reports; the company appears to view high-end displays as a growth engine. (They released 55-inch, 66-inch and 77-inch OLED models earlier in the year, said Yonhap News Agency.) The showing also indicates that LG Display continues to focus attention on OLED.Reisinger offered reasons for why OLED "is widely believed to be the next frontier." He said, "The technology adds an organic compound layer that allows not only for exceedingly thin screens, but for those displays to be curved. The organic material also emits its own light, eliminating the need for a backlight. That allows for such thin screens and has made OLED a desirable choice not only for televisions, but for a wide range of wearables and other mobile products."While the wall-sticking panel is a delight to view, Ryan Waniata, writing in Digital Trends on Tuesday, expressed his view that "Such a display probably won't be used in a TV anytime in the near future; it's more likely to end up in wearable technology, automobile manufacturing, and commercial applications." Still, he added, "we could conceivably see such technology (paired with an outboard processing unit) becoming the TV of the future."