Wing-kei Yu

SRAM-DRAM hybrid memory with applications to efficient register files in fine-grained multi-threading

Large register files are common in highly multi-threaded architectures such as GPUs. This paper presents a hybrid memory design that tightly integrates embedded DRAM into SRAM cells with a main application to reducing area and power consumption of multi-threaded register files. In the hybrid memory, each SRAM cell is augmented with multiple DRAM cells so that multiple bits can be stored in each cell. This configuration results in significant area and energy savings compared to the SRAM array with the same capacity due to compact DRAM cells. On other hand, the hybrid memory requires explicit data movements in order to access DRAM contexts. In order to minimize context switching impact, we introduce write-back buffers, background context switching, and context-aware thread scheduling, to the processor pipeline and the scheduler. Circuit and architecture simulations of GPU benchmarks suites show significant savings in register file area (38%) and energy (68%) over the traditional SRAM implementation, with minimal (1.4%) performance loss.

Flash Memory for Ubiquitous Hardware Security Functions: True Random Number Generation and Device Fingerprints

We demonstrate that unmodified commercial Flash memory can provide two important security functions: true random number generation and digital fingerprinting. Taking advantage of random telegraph noise (a type of quantum noise source in highly scaled Flash memory cells) enables high quality true random number generation at a rate up to 10Kbits / second. A scheme based on partial programming exploits process variation in threshold voltages to allow quick generation of many unique fingerprints that can be used for identification and authentication. Both schemes require no change to Flash chips or interfaces, and do not require additional hardware.

A non-volatile microcontroller with integrated floating-gate transistors

We present a non-volatile processor architecture where its entire state can be almost instantly stored and restored in a non-volatile fashion. This capability is attractive for embedded or mobile devices in highly energy constrained environments. The non-volatile microprocessor can enable long computations to continue across power interruptions on self-powered devices or save idle power consumption without sacrificing responsiveness. To realize this vision, a microprocessor must be able to copy state between volatile and non-volatile storage with minimal latency and energy consumption. Our non-volatile architecture addresses this challenge through a per-cell integration of floating-gate non-volatile transistors into volatile state elements and careful system-level optimizations to hide expensive non-volatile operations. We evaluate this approach with a transistor-level prototype of an 8-bit nonvolatile microcontroller. Experiments indicate that the proposed architecture has minimal impact on normal operation while enabling all processor state to be preserved across an unexpected power interruption.

Extracting device fingerprints from flash memory by exploiting physical variations

We evaluate seven techniques for extracting unique signatures from NAND flash devices based on observable effects of process variation. Four of the techniques yield usable signatures that represent different trade-offs between speed, robustness, randomness, and wear imposed on the flash device. We describe how to use the signatures to prevent counterfeiting and uniquely identify and/or authenticate electronic devices.

Hiding Information in Flash Memory

This paper introduces a novel information hiding technique for Flash memory. The method hides data within an analog characteristic of Flash, the program time of individual bits. Because the technique uses analog behaviors, normal Flash memory operations are not affected and hidden information is invisible in the data stored in the memory. Even if an attacker checks a Flash chips analog characteristics, experimental results indicate that the hidden information is difficult to distinguish from inherent manufacturing variation or normal wear on the device. Moreover, the hidden data can survive erasure of the Flash memory data, and the technique can be used on current Flash chips without hardware changes.

Electron-impact ionization of CCl4 and CCl2F2

Absolute partial and total cross sections for electron-impact ionization of CCl4 and CCl2F2 are reported for electron energies from threshold to 1000 eV. The product ions are mass analyzed using a time-of-flight mass spectrometer and detected with a position-sensitive detector whose output demonstrates that all product ion species are collected with equal efficiency irrespective of their initial kinetic energies. Data are presented for production of CCl3(+), CCl2(+), CCl+, C+, Cl2(+), and CCl3(2+) from CCl4; and for production of CCl(2)F+, CClF2(+), CClF(+), (CCl+ + CF2(+)), Cl+, CF+, F+, and C+ from CCl2F2. Data are also reported for formation of (CCl2(+),Cl+) and (CCl+, Cl+) ion pairs from CCl4. The total cross section for each target is obtained as the sum of the partial cross sections. The overall uncertainty in the absolute cross sections for most of the singly charged ions is +/- 5-7 %. The present partial cross sections for lighter fragment ions are found to be considerably greater than had been previously reported but the most recent total cross section measurements agree well with those reported here. Neither the binary-encounter-Bethe theory nor the Deutsch-Mark theory reproduces the experimental cross sections correctly for both targets.

Understanding sources of variations in flash memory for physical unclonable functions

This paper provides detailed characterizations of physical sources behind Flash memory based Physical Unclonable Functions (FPUFs). Universal process variations in Flash physical systems are identified and decomposed into layout, intrinsic, stress and bit-wise fluctuation sources. The study shows the understanding of systematic variations and noise sources are essential for improving the security and reliability of FPUFs. Bitwise variations are proven to be originated mainly from random dopant fluctuation, which is indeed truly random and impossible to clone. Overall, this paper provides a theoretical foundation for the security of FPUFs whereas previous PUF studies rely only on experimental evidence for its security and entropy.

Low power nonvolatile SRAM circuit with integrated low voltage nanocrystal PMOS Flash

This paper presents a new nonvolatile SRAM design that incorporates low-voltage nanocrystal PMOS Flash transistors. The design enables global store, restore and erase operations with negligible penalty on regular SRAM operation. Store/erase operations also do not consume much power even considering charge pump circuits. Circuit simulations based on experimental I–V characteristics demonstrate that 10 µs store/erase operation at ± 6 Vis sufficient for correct restoration of the stored bit even under reasonable process variation.

Electron transfer in collisions of keV hydrogen atoms and ions with methane

Absolute differential cross sections are reported for electron capture and loss by 1–5 keV H atoms incident on CH4 for laboratory scattering angles up to 1.62°, and for charge transfer of 0.5–5 keV H+ with CH4 for scattering angles up to 2.09°. Electron-loss collisions are seen to result in comparatively large scattering angles and a very clear similarity exists between the present differential cross sections and those reported for other molecular targets. The present charge-transfer differential cross sections are consistent with that of Gao et al (1990 Phys. Rev. A 41 5929–33) but not with the calculations of Kimura et al (1995 Phys. Rev. A 52 1196–205). Prior experimental studies of electron-loss and charge-transfer are generally in good accord with the integral values reported here as are the calculations of Kusakabe et al (2000 Phys. Rev. A 62 062715).

Ionization of CCl4 and CCl2F2 by Electron Impact

Absolute partial and total cross sections for electron‐impact ionization of CCl4 and CCl2F2 are reported for electron energies from threshold to 1000 eV. The product ions are mass analyzed using a time‐of‐flight mass spectrometer and detected with a position‐sensitive detector whose output demonstrates that all product ion species are collected with equal efficiency irrespective of their initial kinetic energies. Data are presented for production of CCl3+, CCl2+, CCl+, C+, Cl2+, and CCl32+ from CCl4; and for production of CCl2F+, CClF+, (CCl+ + CF2+), Cl+, CF+, F+, and C+ from CCl2F2. Data are also reported for formation of (CCl2+, Cl+) and (CCl+, Cl+) ion pairs from CCl4. The present partial cross sections for lighter fragment ions are found to be considerably greater than had been previously reported but the most recent total cross section measurements agree well with those reported here.