Leakage Power Reduction Research Articles

An efficient leakage power optimization framework based on reinforcement learning with graph neural network

Threshold voltage (Vth\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$V_{th}$$\\end{document}) assignment is convenient for leakage optimization due to the exponential relation between leakage power and Vth\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$V_{th}$$\\end{document} by swapping logic cells without routing effort. However, it poses great challenge in large scale circuit design as an NP-hard problem. Machine learning-based approaches have been proposed to solve this problem, aiming to achieve well tradeoff between leakage power reduction and runtime speed up without new induced timing violation. In this paper, a leakage power optimization framework based on reinforcement learning (RL) with graph neural network (GNN) is first-ever proposed to formulate Vth\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$V_{th}$$\\end{document} assignment as a RL process by learning timing and physical characteristics of each circuit instance with GNN. Multiple instances are selected in a non-overlapped manner for each RL action iteration to speed up convergence and decouple timing interdependence along circuit path, where the corresponding reward is carefully defined to tradeoff between leakage reduction and potential timing violation. The proposed framework was validated by the Opencores and IWLS 2005 benchmark circuits with TSMC 28 nm technology. Experimental results demonstrate that our work outperforms prior non-analytical and GNN-based methods with better leakage power optimization by additional 5% to 17% reduction, which is highly consistent with the commercial tool. When transferring the trained RL-based framework to unseen circuits, it achieves the roughly identical leakage optimization results as seen circuit and speed up the runtime by 5.7×\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ imes$$\\end{document} to 8.5×\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\ imes$$\\end{document} compared with commercial tool.

Analysis of etched drain based Cylindrical agate‐all‐around tunnel field effect transistor based static random access memory cell design

AbstractThis paper aims to propose a novel method for designing an static random access memory (SRAM) cell using an etched drain based Cyl GAA TFET with a hetero‐substrate material and an elevated density strip. The aim is to reduce power dissipation and improve stability, as demonstrated through analysis utilizing static noise margin (SNM) as well as N‐curve methods. With respect to the 16 nm MOSFET based SRAM cell, the proposed device‐based SRAM cell shows significant improvements with a 68.305% reduction in leakage power, a 15.58% increase in static voltage noise margin (SVNM), an 8.623% increase in static current noise margin (SINM), an 8.152% increase in write trip voltage (WTV), a 12.86% increase in write trip current (WTI), a 27.62% increase in static power noise margin (SPNM), and a 19.95% increase in write trip power (WTP). The design is implemented and analyzed using Cadence Virtuoso software, and a novel approach of look up tables and Verilog A is utilized for the device to circuit application. These results indicate promising advancements in the design of SRAM cells, which could have significant implications for the development of advanced computer systems.

A FinFET Based Low-Power Write Enhanced SRAM Cell With Improved Stability

This work introduces a FinFet based low-power 11T SRAM cell with write enhanced feature, which is considered to meet modern technology requirements due to its distinctive features and performance. The proposed 11T SRAM cell is designed in such a way, so that it reduces the overall power consumption, improving access time, and static noise margin (SNM), especially during the write operation. The proposed 11T SRAM cell is compared with other SRAM cells, including conventional 6T (conv6T), dual pull-up transistor 10T (DPUT10T), dual pull-down transistor 10T (DPDT10T), dual transmission gate 9T (DTG9T), and dual transmission gate 10T (DTG10T), at an equitable standard. The results obtained at the supply voltage of 0.5 V @ 27 °C shows the reduction in leakage power during hold 1 operation by 1.61×/1.33×/1.44×/1.63×/1.46×, and reduction in power consumption during read operation by 1.02×/1.55×/ 1.01×/1.81×/1.97× in comparison with conv6T/DPUT10T/ DPDT10T/DTG9T/DTG10T, respectively. Write access time is also improved by a factor of 1.67×/1.71×/2.59×/1.45×/1.97×, respectively. Additionally, read static noise margin (RSNM) and write static noise margin (WSNM) are improved by 2.03×/1.01×/1.65×/1.54×/1.43×, and 1.42×/1.33×/1.41×/1.02×/1.62×, respectively. This demonstrates why the proposed low-power 11T SRAM cell is desirable, especially when compared to the other cells under consideration.

Accelerating Linear Congruential Generators with Carbon Nanotube Field-Effect Transistors

Abstract In the era of technological advancements, safeguarding information privacy has become paramount across sectors like defense, finance, healthcare, and more. At the heart of these security measures lies cryptography, a field that necessitates the efficient processing and secure storage of data. Our project focuses on the optimization of Linear Congruential Generators (LCG), pivotal for cryptographic applications and pseudorandom number generation. Here it has been pioneered the implementation of LCG using Carbon Nanotube Field-Effect Transistor (CNTFET) technology, targeting the reduction of leakage power through the innovative application of sleep techniques. This advanced approach not only embodies a leap towards minimizing energy consumption but also enhances the overall performance and reliability of cryptographic systems. By integrating sleep modes into the CNTFET-based LCG design, our methodology significantly curtails leakage power without compromising the generator’s efficiency or output quality. The successful implementation and synthesis of this design, employing cutting-edge CNTFET technology, underscore the potential for substantial improvements in power efficiency while maintaining high standards of security and randomness essential for cryptographic algorithms.

FinFET-Based Low-Power Improved PDP 4:2 Approximate Compressor Design

In the era of low-power very large-scale integration design, approximate computing is a soaring design paradigm that assures energy-efficient circuit design at the cost of some accuracy. Multiplication is a very important operation of an arithmetic unit and compressor is the inherent part of the multiplier. In this paper, an energy-efficient design of a 4:2 approximate compressor is proposed. The inclusion of the extra leakage control transistors in the logic has led to leakage power reduction, which ultimately reduces the power dissipation in the circuits. The proposed 4:2 compressor is simulated at 10[Formula: see text]nm fin-field effect transistor (FinFET) technology. The proposed compressor has been verified for its functionality and robustness. The proposed design shows a power dissipation of 154.48[Formula: see text]nW and a delay of 9.16[Formula: see text]ps. In comparison to other 4:2 approximate compressors, the proposed compressor shows the least power dissipation, the lowest power delay product (PDP) and the highest robustness. The power dissipation of the proposed compressor is 92.84% less than that of the base compressor. The PDP of the designed compressor is 80.58% less than the minimum PDP among other compressors. The aging analysis for the negative bias temperature instability (NBTI) effect is also checked for the proposed compressor and compared with the existing designs.

A CNTFET based stable, single-ended 7T SRAM cell with improved write operation

Many researchers are working to improve the write operation in SRAM bit-cell for better write stability, low power dissipation, and minimal access time during the write process. However, the read and hold operation parameters should not be compromised to achieve these improvements. This paper presents a stable single-ended seven-carbon nanotube field-effect transistor (CNTFET) driven SRAM cell with improved write operation. The one-side inverter weakening approach for write and transistor decoupling for read operation leads to reduced dynamic power, low write delay, reduced leakage power, and improved stability. The proposed design is compared with conventional 6T (Conv6T) and three recently proposed designs, i.e., feedback-cutting 8T (feed-cut 8T), Low-power 8T and low-leakage 7T cell. The write delay and write PDP of the proposed design improve by 4.05×/3.58×/1.19×/1.21×and 11.11×/24.71×/2.96×/3.32×, respectively, compared to Conv6T/feed-cut 8T/ low-power 8T/ low-leakage 7T. Also, the read delay and read PDP of the proposed design improve by 1×/1.03×/1.72×/1.56× and 1×/1.03×/1.82×/1.77×, respectively, compared to Conv6T/feed-cut 8T/ low-power 8T/ low-leakage 7T. The leakage power of the proposed design is reduced by 1.08×/1.84×/0.46×/0.72× compared to Conv6T/feed-cut 8T/ low-power 8T/ low-leakage 7T. The noise margin of the proposed cell for hold/write/read operation is improved by 1.02×/1.05×/0.99×compared to the Conv6T design. The simulation was performed using Stanford University’s 32 nm CNTFET model on the cadence virtuoso platform.

A FinFET-based low-power, stable 8T SRAM cell with high yield

Modern battery-enabled systems, such as IoT, require SRAM cells that can maintain data and respond quickly to requests. However, achieving low-power and stable SRAM cells, which are essential for IoT systems, is not possible with CMOS technology due to scaling issues. As a result, new nanoscale devices, such as FinFET, have been introduced as a potential replacement for CMOS in SRAM design. This paper presents a new FinFET-based 8 T SRAM cell with separate reading and writing paths. The cell uses read-decoupling and write-assist pull-down path cut techniques to improve read stability and writability, respectively. Single-ended reading and writing structures reduce power consumption, and stacking transistors, along with read bitline leakage elimination, reduces leakage power dissipation. The proposed cell’s efficiency is evaluated using HSPICE simulator with 10-nm FinFET technology and is compared with conventional 6 T, single-bitline 7 T (SB7T), single-ended 8 T (SE8T), and transmission gate read-decoupled 9 T (TRD9T) SRAM cells at VDD = 0.35 V. The proposed cell improves read stability by 2.59×/1.04 × and enhances writability by 1.38×/1.01 × compared to 6 T/TRD9T and 6 T/SB7T, respectively. It also reduces read power by 28.25 %/19.60 % compared to the 6 T/SB7T and reduces write power by at most 33.40 % and at least 5.46 %. Moreover, the leakage power is reduced by at least 5.80 %. However, the proposed cell increases read/write delay by 1.77×/1.40 × and shows 1.23 × larger layout area compared to 6 T.

Adiabatic Power Management Techniques for Metal Oxide Silicon Field Effect Transistors in Digital Logic Circuits

Metal Oxide Silicon Field Effect Transistors are foundational components in electronics, mainly digital logic circuits. Effective power management in MOSFET-based logic circuits becomes more crucial as technology expands to sub-100-nanometer nodes. Four optimization techniques to reduce power losses during switching transitions are carefully examined in this article. With the first method, QSSERL, no additional timing control clocks are required, which results in significant energy savings, especially in high-frequency applications. In the second method, DFAL, split-level sinusoidal power clocks are used instead of diodes, significantly lowering voltage differentials and power dissipation. Building on DFAL, IDFAL employs control transistors to reduce leakage power, significantly lowering power dissipation and facilitating charge recovery. Lastly, FSOA, inspired by natural species' collective behaviour, enhances key components within MOSFET-based circuits. This paper examines numerous adiabatic logic topics. Developing energy-efficient MOSFET-based logic circuits is essential for improving electronics and satisfying the growing demand for energy-conscious technologies.

Design of a 0.4 V, 8.43 ENOB, 5.29 nW, 2 kS/s SAR ADC for Implantable Devices

This paper presents a 9-bit differential, minimum-powered, successive approximation register (SAR) ADC intended for implantable devices or sensors. Such applications demand nanowatt-range power consumption, which is achieved by designing the SAR ADC with a proposed bootstrap switch, bespoke split-capacitive DAC, customized comparator and a modified dynamic bit-slice unit for SAR logic. The linearity of the ADC is improved by introducing a bootstrap switch with a low clock feedthrough and threshold voltage variations along with the disseminated attenuation capacitor in the split-capacitive DAC. The dynamic comparator is customized to be simple in terms of the number of transistors to gain the advantage of low power and is also designed to have a low dynamic offset voltage. The stacking concept is embedded in the bit-slice unit of SAR logic to achieve reduced leakage power. This paper is concerned with how to contribute to low power consumption in all the aspects possible related to the implementation of the SAR ADC. With a 0.4 V supply and at 2 kS/s, the proposed ADC achieves an SNDR of 52.52 dB and a power consumption of 5.29 nW, resulting in a figure of merit (FOM) of 7.66 fJ/conversion-step.

Design of Enhanced Reversible 9T SRAM Design for the Reduction in Sub-threshold Leakage Current with14nm FinFET Technology

Power dissipation is considered one of the important issues in low power Very-large-scale integration (VLSI) circuit design and is related to the threshold voltage. Generally, the sub-threshold leakage current and the leakage power dissipation are increased by reducing the threshold voltage. The overall performance of the circuit completely depends on this leakage power dissipation because this leakage and power consumption causes the components that are functioning by the battery for a long period to be washed-out rapidly. In this research, the reversible logic gate-based 9T static random access memory (SRAM) is designed in 14nm FinFET technology to reduce leakage power consumption in memory related applications. The Schmitt-trigger (ST)-based 9T SRAM cell is designed to attain high read-write stability and low power consumption using a single bit line structure. The reversible logic gates of Feynman (FG) and Fredkin gate (FRG) are combined to develop a row and column decoder in an SRAM design to diminish the leakage power. Moreover, the transistor stacking effect is applied to the proposed memory design to reduce the leakage power in active mode. The proposed reversible logic and transistor stacking based SRAM design is implemented in Tanner EDA Tool version 16.0. It also performs both read and write operations using the proposed circuit. The performance measures of read access time (RAT), write access time (WAT), read, write, and static power by varying supply voltage and temperature, delay and stability analysis (read/write static noise margin) are examined and compared with existing SRAM designs.

Design of high efficient low power static logic circuit using SG FinFET

ABSTRACT The increasing demand of integration density improvement and battery-powered device efficiency reduced complementary metal-oxide semiconductor (CMOS) technology node. In the technology of CMOS, the components are mainly affected with leakage power, dynamic switching power, short circuit power, gate oxide tunneling leakage current, sub threshold leakage current, and so on. To reduce the above limitations, design of high efficient low power static logic circuit using shorted-gate (SG) fin field-effect transistor (FinFET) based INput DEPendent (INDEP) in 22 nm CMOS technology (SLC-SG-FinFET-INDEP-22 nm CMOS) approach is proposed in this manuscript. The better selection of inputs to proposed INDEP FinFETs model is used for reducing leakage power. The efficiency of the proposed SLC-SG-FinFET-INDEP-22 nm CMOS technique is analysed using delay, power, power delay product, and stability analysis using noise margin. Thus, the proposed SLC-SG-FinFET-INDEP-22 nm CMOS has attained 21.31%, 41.47% and 12.7% lower delay, 20.87%, 34.5% and 22.41% lower power and 4.5%, 25.7% and 32.11% higher speed than existing methods static logic circuit input-controlled leakage restrainer transistor in 22 nm CMOS (SLC-ICLRT-22 nm CMOS), static logic circuit using self-control leakage-suppression block in 22 nm CMOS technology (SLC-SCLSB-22 nm CMOS), and static logic circuit using computational digital low dropout in 22 nm CMOS technology (SLC-CDLDO-22 nm CMOS) methods, respectively.

Enhanced Quasi-Static Energy Recovery and Modified Clock Gating for the Wallace Tree Multiplier with Reversible Adders

High-speed, low-power digital circuits are essential in many digital applications. The objective of the paper is to design high-speed low-power multipliers with reversible gates. The realization of a multiplier with an efficient clocking system is also presented to reduce leakage power consumption. The existing reversible configurations do not possess an effective clocking system for reducing power in an idle state. In this work, a modified clock gating (MCG) and Enhanced Quasi-Static Energy Recovery Logic (EQSERL) are designed to generate a clock pulse. The EQSERL reduces the dynamic state in the circuit and it is designed with reduced complexity. This process effectively reduces leakage power consumption in the circuit. The conventional clock gating circuit is modified for connecting it with the EQSERL. This MCG circuit reduces unwanted power consumption. The conventional Wallace tree multiplier (WTM) involves full adder and half adder circuits. In this proposed work, a square root carry select adder (SQRT CSLA) with a reversible binary to excess-1 converter (BEC) and a ripple carry adder is presented. The reversible gates such as the Feynman gate, the Toffoli gate, and the dual key gate are employed. Finally, the proposed architecture is tested with different clocking circuits using Synopsys tools and the performance metrics like delay and power are observed.

Review of the Nanoscale FinFET Device for the Applications in Nano-regime

Background:The insatiable need for low-power and high-performance integrated circuit (IC) results in the development of alternative options for metal oxide semiconductor field effect transistor (MOSFET) in the ultra-nanoscale regime. The practical challenge of the device scaling limits the use of MOSFET for future technology nodes. ICs are equipped with billions of transistors whose size must be scaled while increasing performance. As the size of the transistor shrinks for the new technology node, the control of the gate over the channel also reduces, leading to sub-threshold leakage. The non-planar technology is the potential methodology to design the ICs for the future technology nodes. The fin-shaped field effect transistor (FinFET) is the most valuable non-planar technology. High sub-threshold slope, better short channel effect (SCE) control, high current drive strength, low dopant-prompted variations, and decreased power dissipation are the prominent features of FinFET technology.Objective:FinFET is an advanced version of MOSFET in terms of geometrical structure. Therefore, in this review paper, the different geometrical structures, working operations, design challenges, future aspects, and the different configurations of FinFETs are presented. The performance of the different configurations of a 1-bit full adder is evaluated and compared.Methods:An overview of FinFET evolution from the planar MOSFET, along with its architecture supported by the requisite equations, is presented in the paper. Besides this, it also gives an insight into the circuit simulation using the FinFETs for the process voltage temperature (PVT) variations, width quantization, design challenges, and the future of FinFETs. A comparative study of FinFET-based 1-bit full adder using various techniques is done to compute and compare the leakage power, delay, and power delay product (PDP).Results:The full adders using FinFETs show less leakage power and PDP. The AND-OR logicbased hybrid full adder using FinFETs shows the least energy consumption per switching. Fin- FET-based gate diffusion input adder shows a 74 % reduction in dynamic power compared to the full adder using MOSFET technology. The low power FinFET-based full adder shows a 54.16 % reduction in leakage power compared to the MOSFET-based full adder. The results signify the effect of multi-gates in curbing the leakage power dissipation.Conclusion:MOSFET faces the practical challenge of device scaling and SCEs at lower technology nodes. It initiates the multi-gate technology for future system generation. FinFET has the capability to design low-power and high-performance circuits in an ultra-nanoscale regime. The geometrical structure of FinFET plays a key role to improve the performance metrics in an ultrananoscale regime.

A low-power SRAM design with enhanced stability and ION/IOFF ratio in FinFET technology for wearable device applications

ABSTRACT Wearable device applications such as smartwatches, fitness trackers, and health monitors rely on batteries for power and require static random-access memory (SRAM) with low power consumption and high stability to ensure accurate sensor readings during prolonged operation. In this regard, this paper proposes a novel low-power single-ended 10-transistor (SE10T) SRAM cell with high stabilities. It utilises a stacked pull-down structure for hold/read stability improvement and leakage power reduction, a feedback-cutting mechanism for writability enhancement, and single-ended read/write structures for dynamic power reduction. Also, using a single-transistor reading path with eliminated read bitline leakage enhances ON-to-OFF currents (I ON /I OFF ) ratio. The proposed design is compared with the conventional 6T, Schmitt-trigger 10T (ST10T), differential writing 10T (DW10T), data-independent read port 10T (DIRP10T), transmission gate read-decoupled 9T (TGRD9T), and feedback-cutting 11T (FC11T) SRAM cells based on 7-nm Fin-shaped Field-Effect Transistor (FinFET) technology at V DD = 0.4 V. The proposed design shows at least 1.04×/1.01×/1.12×/1.25×/1.15× improvement in read stability/writability/read delay/leakage power/dynamic power. Moreover, it shows at least 1.22× improvement in I ON /I OFF ratio and offers the lowest minimum operating voltage (0.292 V). However, it offers a 1.02× lower hold stability than that of ST10T, a 1.18×/1.14×/1.13× higher write delay compared to 6T/ST10T/DW10T, and a 1.56×/1.07×/1.05× higher layout area occupation in comparison with 6T/TGRD9T/DIRP10T. Therefore, the proposed SRAM cell can be an optimum candidate for usage in smartwatches.

Design and Simulation of a Novel 16T SRAM Cell for Low Power Memory Architecture

Static random access memory (SRAM) is a sort of RAM where information is not permanently stored and does not require routine updating. To reduce leakage power without sacrificing performance, a variety of approaches have been applied to SRAM cells. In this study, we suggest a new 16T SRAM design that can function in active, park, standby, or hold modes. A fully static mode of operation for SRAM is made possible by the 16T SRAM structure, which also enhances write margin (WM) and removes charging conflict between devices during read and write operations. The key objectives of the suggested architecture are to retain logic state in park mode while maintaining stability and reducing standby time in active mode, as well as to reduce leakage current in standby mode. This is done by removing feedback from the back-to-back inverters during write operations via the data-dependent supply block. This makes it possible for the suggested bitcell to considerably increase the WM. A novel SRAM cell with 16 transistors is created with subthreshold operation and enhanced data stability. By using an equalized bit line technique to avoid leakage due to enhanced data pattern and RBL detection, the suggested single-ended SRAM cell with dynamic feedback control lowers the static noise margin for ultra-low power transfer. A sleep transistor is incorporated into the architecture to save power consumption when the system is in standby mode due to inefficient voltage transmission. Tanner EDA tool V.14.1 on 45-nm CMOS was used for design and simulation. The outcomes demonstrate a considerable decrease in no-load current and power loss.

Novel PVT Resilient Low-Power Dynamic XOR/XNOR Design Using Variable Threshold MOS for IoT Applications

A variable threshold voltage hybrid evaluation network based dynamic XOR/XNOR gate is presented to reduce the parameters leakage power dissipation, dynamic power, and layout area compared to existing dynamic gates under similar delay time conditions. This study explores the impacts of process, supply voltage, and temperature changes on leakage power dissipation and dynamic power using Monte Carlo simulation. The Monte Carlo analysis demonstrates that leakage power dissipation and dynamic power reduction have significantly improved. Furthermore, when compared to hybrid type dynamic XOR/XNOR (N-type XOR/XNOR), the proposed design reduces leakage power and dynamic power consumption by 6.1% (54.0%) and 18.75% (35.0%), respectively. While the proposed design has a slight layout area penalty compared to a hybrid type dynamic XOR/XNOR, it offers the same amount of the layout area as a traditional N-type XOR/XNOR.

Accurate Estimation of Data Retention Voltage (DRV) for a 6T SRAM Cell at 45 nm, 65 nm, and 130 nm Technology Nodes

Scaling of supply voltage is one of the well-known methods for reducing leakage power. Data Retention Voltage (DRV) is the lower limit for the scaled voltage of SRAM. In this paper, we propose a modified analytical model for estimating the DRV of a 6T SRAM cell. The model is based on the sub-threshold current equations of the SRAM cell. The DRV of the 6T SRAM cell is computed using this model which is verified with the simulations carried out using the Cadence Virtuoso Tool. GPDK 45 nm, UMC 130 and 65 nm technology files are used for the validation of the model. The simulations are carried out for the 6T cell with different cell ratios and pull-up ratios. For each case, the DRV estimated from the model is nearly equal to the expected DRV calculated from simulations. The results matched well with the simulated results around 93%–79% whereas Qin’s model which matches only 75%–74%. This model shows better results than Qin’s model in terms of complexity and computation time. Furthermore, the proposed model is also validated for the calculation of DRV during process variation and shows better results.

Improving Energy Efficiency of CGRAs with Low-Overhead Fine-Grained Power Domains

To effectively minimize static power for a wide range of applications, power domains for coarse-grained reconfigurable array (CGRA) architectures need to be more fine-grained than those found in a typical application-specific integrated circuit. However, the special isolation logic needed to ensure electrical protection between off and on domains makes fine-grained power domains area- and timing-inefficient. We propose a novel design of the CGRA routing fabric that reduces the area overhead of power domain boundary protection from around 9% to less than 1% without incurring any extra timing delay from the isolation cells. Conventional Unified Power Format based flow for power domain boundary protection does not support this design choice. Therefore, we create our own compiler-like passes that iteratively introduce the needed design changes, and formally verify the transformations using methods based on satisfiability modulo theories. These passes also let us optimize how we handle test and debug signals through the off tiles in the CGRA. Using our framework, we add power domains to a CGRA that we designed and taped out. The CGRA has 32 × 16 processing element and memory tiles and 4-MB secondary memory. We address the implementation challenges encountered due to the introduction of fine-grained power domains, including the addressing of the CGRA tiles, the power grid design, well substrate connections, and distribution of global signals. Our CGRA achieves up to 83% reduction in leakage power and 26% reduction in total power versus an identical CGRA without multiple power domains, for a range of image processing and machine learning applications.

Dynamic power-gating for leakage power reduction in FPGAs

Dynamic power-gating for leakage power reduction in FPGAs

A Variation-Aware MTJ Store Energy Estimation Model for Edge Devices With Verify-and-Retryable Nonvolatile Flip-Flops

While the spin-transfer torque (STT) magnetic tunnel junction (MTJ) is a promising technique for enabling nonvolatile flip-flops (NVFFs) to perform power gating to reduce leakage power without any data losses, the large store energy (the energy to make a store operation) of MTJs needs to be addressed. The nonvolatile cool mega array series is an edge-oriented coarse-grained reconfigurable accelerator that implements an improved MTJ-based NVFF with a verify-and-retryable store method that should ideally reduce the store energy under the presence of the switching time variation originating from the stochastic nature of the MTJs. However, the energy reduction effect of the method has not been formulated or evaluated thoroughly enough to make the best use of the method in actual applications. In this study, we propose an analytical model to estimate the store energy in typical operational conditions under the assumption of switching time variations following the normal distributions based on the measurements of a real chip fabricated with a 40-nm perpendicular MTJ/CMOS hybrid process. In contrast to the tedious measurement on each different condition, the proposed model allows for an instantaneous determination of the best storing method for minimizing the store energy, with an energy reduction of up to 69% compared with a conventional one-time attempt storing method. This model is expected to be used for system-level energy simulations and, ultimately, for design explorations in pursuit of energy-optimized memory.