First new cache-coherence mechanism in 30 years

More efficient memory-management scheme could help enable chips with thousands of cores.

In a modern, multicore chip, every core — or processor — has its own small memory cache, where it stores frequently used data. But the chip also has a larger, shared cache, which all the cores can access.

If one core tries to update data in the shared cache, other cores working on the same data need to know. So the shared cache keeps a directory of which cores have copies of which data.

That directory takes up a significant chunk of memory: In a 64-core chip, it might be 12 percent of the shared cache. And that percentage will only increase with the core count. Envisioned chips with 128, 256, or even 1,000 cores will need a more efficient way of maintaining cache coherence.

At the International Conference on Parallel Architectures and Compilation Techniques in October, MIT researchers unveil the first fundamentally new approach to cache coherence in more than three decades. Whereas with existing techniques, the directory’s memory allotment increases in direct proportion to the number of cores, with the new approach, it increases according to the logarithm of the number of cores.

In a 128-core chip, that means that the new technique would require only one-third as much memory as its predecessor. With Intel set to release a 72-core high-performance chip in the near future, that’s a more than hypothetical advantage. But with a 256-core chip, the space savings rises to 80 percent, and with a 1,000-core chip, 96 percent.

When multiple cores are simply reading data stored at the same location, there’s no problem. Conflicts arise only when one of the cores needs to update the shared data. With a directory system, the chip looks up which cores are working on that data and sends them messages invalidating their locally stored copies of it.

“Directories guarantee that when a write happens, no stale copies of the data exist,” says Xiangyao Yu, an MIT graduate student in electrical engineering and computer science and first author on the new paper. “After this write happens, no read to the previous version should happen. So this write is ordered after all the previous reads in physical-time order.”

Time travel

What Yu and his thesis advisor — Srini Devadas, the Edwin Sibley Webster Professor in MIT’s Department of Electrical Engineering and Computer Science — realized was that the physical-time order of distributed computations doesn’t really matter, so long as their logical-time order is preserved. That is, core A can keep working away on a piece of data that core B has since overwritten, provided that the rest of the system treats core A’s work as having preceded core B’s.

The ingenuity of Yu and Devadas’ approach is in finding a simple and efficient means of enforcing a global logical-time ordering. “What we do is we just assign time stamps to each operation, and we make sure that all the operations follow that time stamp order,” Yu says.

With Yu and Devadas’ system, each core has its own counter, and each data item in memory has an associated counter, too. When a program launches, all the counters are set to zero. When a core reads a piece of data, it takes out a “lease” on it, meaning that it increments the data item’s counter to, say, 10. As long as the core’s internal counter doesn’t exceed 10, its copy of the data is valid. (The particular numbers don’t matter much; what matters is their relative value.)

When a core needs to overwrite the data, however, it takes “ownership” of it. Other cores can continue working on their locally stored copies of the data, but if they want to extend their leases, they have to coordinate with the data item’s owner. The core that’s doing the writing increments its internal counter to a value that’s higher than the last value of the data item’s counter.

Say, for instance, that cores A through D have all read the same data, setting their internal counters to 1 and incrementing the data’s counter to 10. Core E needs to overwrite the data, so it takes ownership of it and sets its internal counter to 11. Its internal counter now designates it as operating at a later logical time than the other cores: They’re way back at 1, and it’s ahead at 11. The idea of leaping forward in time is what gives the system its name — Tardis, after the time-traveling spaceship of the British science fiction hero Dr. Who.

Now, if core A tries to take out a new lease on the data, it will find it owned by core E, to which it sends a message. Core E writes the data back to the shared cache, and core A reads it, incrementing its internal counter to 11 or higher.

Unexplored potential

In addition to saving space in memory, Tardis also eliminates the need to broadcast invalidation messages to all the cores that are sharing a data item. In massively multicore chips, Yu says, this could lead to performance improvements as well. “We didn’t see performance gains from that in these experiments,” Yu says. “But that may depend on the benchmarks” — the industry-standard programs on which Yu and Devadas tested Tardis. “They’re highly optimized, so maybe they already removed this bottleneck,” Yu says.

“There have been other people who have looked at this sort of lease idea,” says Christopher Hughes, a principal engineer at Intel Labs, “but at least to my knowledge, they tend to use physical time. You would give a lease to somebody and say, ‘OK, yes, you can use this data for, say, 100 cycles, and I guarantee that nobody else is going to touch it in that amount of time.’ But then you’re kind of capping your performance, because if somebody else immediately afterward wants to change the data, then they’ve got to wait 100 cycles before they can do so. Whereas here, no problem, you can just advance the clock. That is something that, to my knowledge, has never been done before. That’s the key idea that’s really neat.”

Hughes says, however, that chip designers are conservative by nature. “Almost all mass-produced commercial systems are based on directory-based protocols,” he says. “We don’t mess with them because it’s so easy to make a mistake when changing the implementation.”

But “part of the advantage of their scheme is that it is conceptually somewhat simpler than current [directory-based] schemes,” he adds. “Another thing that these guys have done is not only propose the idea, but they have a separate paper actually proving its correctness. That’s very important for folks in this field.”

MIT

Ultra-small and Ultra–fast Electro-optic Modulator

Due to the voltage applied, a beam of light (top left) is modulated
by the digital bits (bottom right) of the converter (yellow). An electrical
signal is converted into an optical signal. (Graphics: A. Melikyan/KIT)

Nature Photonics Magazine Presents a World-record Micrometer-sized Converter of Electrical into Optical Signals / Future Energy-efficient Chip-to-chip Optical Communication Links

Thanks to optical signals, mails and data can be transmitted rapidly around the globe. But also exchange of digital information between electronic chips may be accelerated and energy efficiency might be increased by using optical signals. However, this would require simple methods to switch from electrical to optical signals. In the Nature Photonics magazine, researchers now present a device of 29 µm in length, which converts signals at a rate of about 40 gigabits per second. It is the most compact high-speed phase modulator in the world. DOI: 10.1038/NPHOTON.2014.9.

“Conversion of electrical into optical signals happens closer to the processor,” Juerg Leuthold says. He coordinated the research project at the Karlsruhe Institute of Technology and has meanwhile moved to the ETH Zurich. “As a result, speed gains are achieved and conduction losses can be prevented. This might reduce energy consumption of the growing information technology.”

The electro-optical converter consists of two parallel gold electrodes of about 29 µm in length, which is one third of the diameter of a human hair. The electrodes are separated by a gap of about one tenth of a micrometer in width. The voltage applied to the electrodes is synchronized with the digital data. The gap is filled with an electro-optical polymer, whose refraction index changes as a function of the applied voltage. “A continuous beam of light from the silicon waveguide excites electromagnetic surface waves, so-called surface plasmons (SP), in the gap,” Argishti Melikyan, KIT, first author of the publication, explains. “As a result of the voltage applied to the polymer, the phase of the SP is modulated. At the end of the device, the modulated SP enter the exit silicon waveguide in the form of a modulated beam of light. In this way, the data bits are encoded in the phase of the light.”

Their recent results revealed that the electro-optic modulator reliably converts data flows of about 40 gigabits per second. It uses the infrared light of 1480 – 1600 nanometers in wavelength usually encountered in the broadband glass fiber network. Even temperatures of up to 85°C do not cause any operation failures. The presented device is the most compact high-speed phase modulator in the world. It can be produced by well-established CMOS fabrication processes. Integration into current chip architectures is hence possible. “The device combines many advantages of other systems, such as a high modulation speed, compact design, and energy efficiency. In the future, plasmonic devices might be used for signal processing in the terahertz range,” says Christian Koos, spokesperson of KIT’s Helmholtz International Research School of Teratronics (HIRST), which focuses on merging photonic and electronic techniques for high-speed signal processing. ”Hundreds of plasmonic modulators might fit on a chip and data rates in the range of terabits per second might be reached.”

Presently, information and communication systems consume about 10 percent of the electricity in Germany. This includes computers and smartphones of individual users as well as servers at large computing centers. As data traffic grows exponentially, new approaches are required to increasing the capacity of such systems and reducing their energy consumption at the same time. Plasmonic components might be of decisive importance in this respect.

The present paper is part of the EU project NAVOLCHI, Nano Scale Disruptive Silicon-Plasmonic Platform for Chip-to-Chip Interconnection. This project is aimed at using the interaction of light and electrons in metal surfaces for the development of novel components for data transmission between chips. “Conventional electric chip-to-chip data transmission reaches its limits,” says the present project coordinator Manfred Kohl, KIT. “NAVOLCHI is about to overcome those limits using optical technology.” It is funded under the 7th Research Framework Programme of the EU and has a budget of EUR 3.4 million.

For more information on the NAVOLCHI project, clickhttp://www.imt.kit.edu/projects/navolchi/ 

High-speed plasmonic phase modulators, A. Melikyan, L. Alloatti, A. Muslija, D. Hillerkuss, P. C. Schindler, J. Li, R. Palmer, D. Korn, S. Muehlbrandt, D. Van Thourhout, B. Chen, R. Dinu, M. Sommer, C. Koos, M. Kohl, W. Freude, and J. Leuthold, nature photonics AOP, DOI: 10.1038/NPHOTON.2014.9

http://www.nature.com/nphoton/index.html 

Karlsruhe Institute of Technology (KIT) is a public corporation according to the legislation of the state of Baden-Württemberg. It fulfills the mission of a university and the mission of a national research center of the Helmholtz Association. Research activities focus on energy, the natural and built environment as well as on society and technology and cover the whole range extending from fundamental aspects to application. With about 9000 employees, including nearly 6000 staff members in the science and education sector, and 24000 students, KIT is one of the biggest research and education institutions in Europe. Work of KIT is based on the knowledge triangle of research, teaching, and innovation.

Source: http://www.kit.edu/visit/pi_2014_14701.php

Computer Chips that Listen to Bacteria

In a study published today in Nature Communications, a research team led by Ken Shepard, professor of electrical engineering and biomedical engineering at Columbia Engineering, and Lars Dietrich, assistant professor of biological sciences at Columbia University, has demonstrated that integrated circuit technology, the basis of modern computers and communications devices, can be used for a most unusual application—the study of signaling in bacterial colonies. They have developed a chip based on complementary metal-oxide-semiconductor (CMOS) technology that enables them to electrochemically image the signaling molecules from these colonies spatially and temporally. In effect, they have developed chips that “listen” to bacteria.

“This is an exciting new application for CMOS technology that will provide new insights into how biofilms form,” says Shepard. “Disrupting biofilm formation has important implications in public health in reducing infection rates.”

The researchers, who include PhD students Daniel Bellin (electrical engineering) and Hassan Sakhtah (biology), say that this is the first time integrated circuits have been used for such an application—imaging small molecules electrochemically in a multicellular structure. While optical microscopy techniques remain paramount for studying biological systems (using photons allows for relatively non-invasive interaction to the biological system being studied), they cannot directly detect critical components of physiology, such as primary metabolism and signaling factors.

The team thought there might be a better way to directly detect small molecules through techniques that employ direct transduction to electrons, without using photos as an intermediary. They made an integrated circuit, a chip that, Shepard says, is an ‘active’ glass slide, a slide that not only forms a solid-support for the bacterial colony but also ‘listens’ to the bacteria as they talk to each other.”

“This is a big step forward,” Dietrich continues. “We describe using this chip to ‘listen in’ on conversations taking place in biofilms, but we are also proposing to use it to interrupt these conversations and thereby disrupt the biofilm. In addition to the pure science implications of these studies, a potential application of this would be to integrate such chips into medical devices that are common sites of biofilm formation, such as catheters, and then use the chips to limit bacterial colonization.”Cells, Dietrich explains, mediate their physiological activities using secreted molecules. The team looked specifically at phenazines, which are secreted metabolites that control gene expression. Their study found that the bacterial colonies produced a phenazine gradient that, they say, is likely to be of physiological significance and contribute to colony morphogenesis.

The next step for the team will be to develop a larger chip that will enable larger colonies to be imaged at higher spatial and temporal resolutions.

“This represents a new and exciting way in which solid-state electronics can be used to study biological systems,” Shepard adds. “This is one of the many emerging ways integrated circuit technology is having impact in biotechnology and the life sciences.”

The study was supported by the National Institutes of Health and the National Science 

Source: http://engineering.columbia.edu/chips-listen-bacteria-0

Liquid biopsy could improve cancer diagnosis and treatment

A microfluidic chip developed at the University of Michigan is among the best at capturing elusive circulating tumor cells from blood—and it can support the cells' growth for further analysis.

The device, believed to be the first to pair these functions, uses the advanced electronics material graphene oxide. In clinics, such a device could one day help doctors diagnose cancers, give more accurate prognoses and test treatment options on cultured cells without subjecting patients to traditional biopsies.

"If we can get these technologies to work, it will advance new cancer drugs and revolutionize the treatment of cancer patients," said Dr. Max Wicha, Max Wicha, M.D., Distinguished Professor of Oncology and director of the U-M Comprehensive Cancer Center and co-author of a paper on the new device, published online this week in Nature Nanotechnology.

"Circulating tumor cells will play a significant role in the early diagnosis of cancer and to help us understand if treatments are working in our cancer patients by serving as a 'liquid' biopsy to assess treatment responses in real time," said co-author Dr. Diane Simeone, the Lazar J. Greenfield Professor of Surgery at the U-M Medical School and director of the Translational Oncology Program.

"Studies of circulating tumor cells will also help us understand the basic biologic mechanisms by which cancer cells metastasize or spread to distant organs—the major cause of death in cancer patients."

Yet these cells aren't living up to their promise in medicine because they are so difficult to separate from a blood sample, the researchers say. In the blood of early-stage cancer patients, they account for less than one in every billion cells, so catching them is tougher than finding the proverbial needle in a haystack.

"I can burn the haystack or use a huge magnet," said Sunitha Nagrath, an assistant professor of chemical engineering, who led the research. "When it comes to circulating tumor cells, they almost look like—feel like—any other blood cell."

An optical microscope reveals a cancer cell attached to the flower pattern.On their microfluidic chip, Nagrath's team grew dense forests of molecular chains, each equipped with an antibody to grab onto cancer cells.

Even after the cells are caught, it's still hard to run a robust analysis on just a handful of them, the researchers say. That's why this demonstration of highly sensitive tumor cell capture, combined with the ability to grow the cells in the same device, is so promising.

Hyeun Joong Yoon, a postdoctoral researcher in the Nagrath lab with a background in electrical engineering, was instrumental in making the microfluidic chip. He started with a silicon base and added a grid of nearly 60,000 flat gold shapes, like four-petaled flowers, each no wider than a strand of hair.

The gold flowers naturally attracted a relatively new material called graphene oxide. These sheets of carbon and oxygen, just a few atoms thick, layered themselves over the gold. This layered formation allowed the team to grow the tumor-cell-catching molecular chains so densely.

"It's almost like each graphene has many nano-arms to capture cells," Nagrath said.

To test the device, the team ran one-milliliter samples of blood through the chip's thin chamber. Even when they had added just three-to-five cancer cells to the 5-10 billion blood cells, the chip was able to capture all of the cells in the sample half the time, with an average of 73 percent over 10 trials.

"That's the highest anybody has shown in the literature for spiking such a low number of cells," Nagrath said.

Cancer cells glow green with fluorescent tags.The team counted the captured cancer cells by tagging them with fluorescent molecules and viewing them through a microscope. These tags made the cancer cells easy to distinguish from accidentally caught blood cells. They also grew breast cancer cells over six days, using an electron microscope to see how they spread across the gold flowers.

"When you have individual cells, the amount of material in each cell is often so small that it's hard to develop molecular assays," Wicha said. "This device allows the cells to be grown into larger quantities so you can do a genetic analysis more easily."

The chip could capture pancreatic, breast and lung cancer cells from patient samples. Nagrath was surprised that the device was able to catch about four tumor cells per milliliter of blood from the lung cancer patients, even though they had the early-stage form of the disease.

Working in a team that comprises both engineers and medical professionals at U-M, Nagrath is optimistic that the new technique could reach clinics in three years.

The paper is titled "Sensitive capture of circulating tumor cells by functionalized graphene oxide nanosheets." The university is pursuing patent protection for the intellectual property and is seeking commercialization partners to help bring the technology to market.

This research is supported by the National Institutes of Health Director's New Innovator Award No. 1DP2OD006672-01.

Source: http://ns.umich.edu/new/multimedia/slideshows/21719-liquid-biopsy-could-improve-cancer-diagnosis-and-treatment

First Self-Assembling Diamond Quantum Devices Unveiled

The ability to self-assemble quantum components on the nanometer scale could revolutionise the future of computing

One of the great goals of applied physics is to make quantum information processing a robust and common technique. To achieve this, physicists will need a simple way of storing and manipulating quantum information, preferably at room temperature.

There is no shortage of possible quantum storage devices but one sits head and shoulders above most others: a nitrogen atom that has replaced a carbon atom in a diamond lattice, an arrangement known as a nitrogen-vacancy centre.

Today, an international team of physicists say they’ve used biological self-assembly techniques to make diamond-based prototypes of the quantum information storage devices of this type. 

 That’s a development that has the potential to profoundly influence the future of computing.

The key to all this is nitrogen-vacancy centres in diamond which behave like single atoms. They can store photons, emit them again and interact with other nitrogen-vacancy centres nearby. In fact, their photon storage ability is legendary, holding them, and the information the carry, for periods stretching to milliseconds. At room temperature.

That’s significantly longer and more robust than other quantum information storage devices.

When two nitrogen vacancy centres are close enough to together to interact, they can also process quantum information. But this is only possible at distances of less than 10 nanometers or so and therein lies the problem.

The conventional way of making nitrogen vacancy centres is to fire nitrogen ions into a diamond crystal and hope they embed themselves into the structure. But that can only be done at a resolution measured in tens of nanometers. So adjacent nitrogen-vacancy centres are always too far apart to interact.

Clearly some other way of bringing these centres together is desperately needed. Enter Andreas Albrecht at the University of Ulm in Germany and a few pals who have come up with a better way to do it using the self-assembling technique of molecular origami.

Their idea is conceptually simple. For some time now, biologists have been creating complicated three dimensional structures out of biomolecules. Their technique is based on the highly selective reactivity of biomolecules such as DNA. These can be designed to fit together and to other things, like self-assembling Lego bricks, to form three dimensional shapes that are nanoscopic in scale.

Albrecht and co simply modified a well known ring-shaped protein called SP1 so that it binds to diamond. In fact, they created 12 binding sites on this ring allowing it to hold six nanodiamonds in hexagonal formation.

They then used a laser to generate nanodiamonds just 5 nanometres across by blasting them off a larger crystal. They placed the resulting crystals in a liquid which they poured onto a layer of the modified SP1 rings.

When the nanodiamonds bonded with the rings, they formed hexagonal structures in which each nanodiamond is just a few nanometres from its neighbours. That’s exactly the kind of distance required for nitrogen vacancy centres to interact.

They go on to show theoretically that this kind of arrangement should allow for some serious quantum information processing.

Of course, there are challenges ahead. One is to repeat this method with nanodiamonds that actually contain nitrogen vacancies and then to experimentally show that they do indeed interact. Clearly, that must be high on the list of future priorities.

In the meantime, this kind of self-assembly is a powerful new approach that is full of promise. Biologists have been able to create a rich tapestry of shapes and structures using molecular origami.

There’s no reason why Albrecht and co can’t do the same to create vastly more complex arrangements of nitrogen vacancy centres.

These might well become the beating hearts of quantum chips of the future.

 Ref: arxiv.org/abs/1301.1871: Self-Assembling Hybrid Diamond-Biological Quantum Devices

Source: http://www.technologyreview.com/view/519231/first-self-assembling-diamond-quantum-devices-unveiled/