Dynamic power consumption is a significant concern in modern integrated circuits. This issue is primarily caused by signal toggling, including unwanted toggles known as glitches. With the number of operations increasing in circuits, glitches can lead to significant additional dynamic power. This paper presents SMART-GPO, a novel framework that efficiently and accurately estimates and reduces glitch power. Our approach samples cycles for accurate glitch estimation, followed by gate-sizing and Vth assignment to optimize glitch power based on sensitivity measurements. We validated SMART-GPO on the Berkeley Out-of-Order Machine (BOOM) and Rocket SoCs with TSMC N28 technology. It achieves a mean absolute percentage error (MAPE) of 2% on glitch power estimation when running power analysis on only 1% simulation cycles. The optimization results demonstrate that our framework reduces glitch power by more than 9%, which outperforms previous approaches substantially.

Adders are the primary components in the data-path logic of a microprocessor, and thus, adder design has been always a critical issue in the very large-scale integration (VLSI) industry. However, it is infeasible for designers to obtain optimal adder architecture by exhaustively running EDA flow due to the extremely large design space. Previous arts have proposed the machine learning-based framework to explore the design space. Nevertheless, they fall into suboptimality due to a two-stage flow of the learning process and less efficient nor effective feature representations of prefix adder structures. In this article, we first integrate a variational graph autoencoder and a neural process (NP) into an end-to-end, multibranch framework, which is termed thegraph neural process. The former performs automatic feature learning of prefix adder structures, whilst the latter one is designed as an alternative to the Gaussian process. Then, we propose a sequential optimization framework with the graph NP as the surrogate model to explore the Pareto-optimal prefix adder structures with tradeoff among Quality-of-Result (QoR) metrics, such as power, area, and delay. The experimental results show that compared with state-of-the-art methodologies, our framework can achieve a much better Pareto frontier in multiple QoR metric spaces with fewer design-flow evaluations.

We propose a methodology to perform process variation-aware device and circuit design using fully physics-based simulations within limited computational resources, without developing a compact model. Machine learning (ML), specifically a support vector regression (SVR) model, has been used. The SVR model has been trained using a dataset of devices simulated a priori, and the accuracy of prediction by the trained SVR model has been demonstrated. To produce a switching time distribution from the trained ML model, we only had to generate the dataset to train and validate the model, which needed ∼500 hours of computation. On the other hand, if 10 6 samples were to be simulated using the same computation resources to generate a switching time distribution from micromagnetic simulations, it would have taken ∼250 days. Spin-transfer-torque random access memory (STTRAM) has been used to demonstrate the method. However, different physical systems may be considered, different ML models can be used for different physical systems and/or different device parameter sets, and similar ends could be achieved by training the ML model using measured device data.

In spite of maturity to the modern electronic design automation (EDA) tools, optimized designs at architectural stage may become suboptimal after going through physical design flow. Adder design has been such a long studied fundamental problem in very large-scale integration industry yet designers cannot achieve optimal solutions by running EDA tools on the set of available prefix adder architectures. In this paper, we enhance a state-of-the-art prefix adder synthesis algorithm to obtain a much wider solution space in architectural domain. On top of that, a machine learning-based design space exploration methodology is applied to predict the Pareto frontier of the adders in physical domain, which is infeasible by exhaustively running EDA tools for innumerable architectural solutions. Considering the high cost of obtaining the true values for learning, an active learning algorithm is proposed to select the representative data during learning process, which uses less labeled data while achieving better quality of Pareto frontier. Experimental results demonstrate that our framework can achieve Pareto frontier of high quality over a wide design space, bridging the gap between architectural and physical designs. Source code and data are available athttps://github.com/yuzhe630/adder-DSE.

Digital circuit physical unclonable function (PUF) has been attracting attentions for the merits of resilience to the environmental and operational variations that analog PUFs suffer from. Existing state-of-the-art digital circuit PUFs, however, are either hybrid of analog-digital circuits which are still under the shadow of vulnerability, or impractical for real-world applications. In this paper, we propose a novel highly nonlinear and secure digital PUF (D-PUF) and the spliced version SD-PUF. The fingerprints are extracted from intentionally induced very large-scale integration interconnect randomness during lithography process, as well as a post-silicon shuffling process. Strongly skewed CMOS latches are used to ensure the immunity against environmental and operational variations. Crucially, a highly nonlinear logic network is proposed to effectively spread and augment any subtle interconnect randomness, which also enables strong resilience against machine learning attacks. On top of it, the expandable architecture of the proposed logic network empowers a novel post-silicon shuffle-splice mechanism, where multiple randomly selected D-PUFs are spliced to be one SD-PUF, pushing the statistical security to a much higher level, while significantly reducing the mask cost per PUF device. It also decouples the trustworthy demands enforced to the foundries or other third party manufacturers. Our proposed PUFs demonstrate close to ideal performance in terms of statistical metrics, including 0 intra-Hamming distance. Various state-of-the-art machine learning models show prediction accuracies almost no better than random guesses when attacking to the proposed PUFs. We also mathematically prove the probability of existence of identical SD-PUF pair is significantly lower than that of D-PUF pair, e.g., such probability of an SD-PUF spliced by 30 D-PUFs is 2.3 × 10-22, which is 19 order magnitude lower than that of D-PUF. Benefited from the proposed shuffle-splice mechanism, the mask cost per SD-PUF is also reduced by 300× than that of D-PUF.

In spite of maturity to the modern electronic design automation (EDA) tools, optimized designs at architectural stage may become sub-optimal after going through physical design flow. Adder design has been such a long studied fundamental problem in VLSI industry yet designers cannot achieve optimal solutions by running EDA tools on the set of available prefix adder architectures. In this paper, we enhance a state-of-the-art prefix adder synthesis algorithm to obtain a much wider solution space in architectural domain. On top of that, a machine learning based design space exploration methodology is applied to predict the Pareto frontier of the adders in physical domain, which is infeasible by exhaustively running EDA tools for innumerable architectural solutions. Experimental results demonstrate that our framework can achieve near-optimal delay vs. power/area Pareto frontier over a wide design space, bridging the gap between architeon the set of available prefix adder architectures. In this paper, we enhance a state-of-the-art prefix adder synthesis algorithm to obtain a much wider solution space in architectural domain. On top of that, a machine learning based design space exploration methodology is applied to predict the Pareto frontier of the adders in physical domain, which is infeasible by exhaustively running EDA tools for innumerable architectural solutions. Experimental results demonstrate that our framework can achieve near-optimal delay vs. power/area Pareto frontier over a wide design space, bridging the gap between architectural andctural and physical designs.

Conventional silicon physical unclonable function (PUF) extracts fingerprints from transistor's analog attributes, which are vulnerable to environmental and operational variations. Recently, digitalized PUF prototypes have emerged to overcome the vulnerability issues, however, the existing prototypes are either hybrid of analog-digital PUFs which are still under the shadow of vulnerability, or impractical for real-world implementation. To address the above limitations, we propose a learning resilient and reliable digital PUF (LRR-DPUF). The fingerprints are extracted from VLSI interconnect geometrical randomness induced by lithography variations. Crucially, we use strongly skewed latches to ensure the immunity against environmental and operational variations. Further, a cross-coupled, highly non-linear logic network is proposed to effectively spread and augment even subtle interconnect randomness, as well as to achieve strong resilience to machine learning attacks. We demonstrate that a 64-bit LRR-DPUF exhibits close to ideal statistical performances, including 0 intra Hamming Distance. We also mathematically prove that each output of the LRR-DPUF follows uniform distribution. Various state-of-the-art machine learning models show almost no better than random prediction accuracies when applied to LRR-DPUF.

Adders are the most fundamental arithmetic units, and often on the timing critical paths of microprocessors. Among various adder configurations, parallel prefix adders provide the best performance vs. power/area trade-off, especially for higher bit-widths. With aggressive technology scaling, the performance of a parallel prefix adder, in addition to the dependence on the logic-level, is determined by wire-length and congestion which can be mitigated by adjusting fan-out. This paper proposes a polynomial-time algorithm to synthesize n bit parallel prefix adders targeting the minimization of the size of the prefix graph with log2n logic level and any arbitrary fan-out restriction. A structure aware prefix node cloning is then applied to the resultant prefix adder solutions to further optimize the size of the prefix graphs. The design space exploration by our approach provides a set of pareto-optimal solutions for delay vs. power trade-off, and these pareto-optimal solutions can be used in high-performance designs instead of picking from a fixed library (Kogge-Stone, Sklansky, etc.). Experimental results demonstrate that our approach: 1) excels highly competitive industry standard Synopsys design compiler adder, regular adders such as Sklansky adder and Kogge-Stone adder, and a highly runtime/memory intensive recent algorithm in 32 nm technology node and 2) improves performance/area over even 64 bit custom designed adders targeting 22 nm technology library and implemented in an industrial high-performance design.

Modern systems-on-a-chip and microprocessors, e.g., those in smart phones and laptops, typically have multiple operating conditions, such as video streaming, Web browsing, standby, and so on. They will have different performance targets and run under different supply voltages. Gate sizing (with threshold voltage assignment) is a fundamental step for power/performance optimization. However, conventional gate sizing algorithms only consider one scenario, e.g., the performance-critical operating condition, which may be over-design for other operating conditions. In addition, reliability has become a prime concern in nanometer designs, and gate sizing has been employed to mitigate aging. However: 1) previous aging-affected delay models do not take into account more than one operating condition to estimate the aging impact and 2) earlier aging aware gate sizing algorithms only consider one operating condition at a time. In this paper, we present a new paradigm of aging aware gate sizing, one-size-fits-all (OSFA), which performs power/performance optimizations across multiple operating conditions. The existing delay model for negative bias temperature instability (NBTI) is extended to take into account multiple operating conditions, and incorporated into our OSFA framework. Based on OSFA, we also adjust the supply voltage targeting overall power optimization. A speed-up heuristic is proposed to scale our OSFA design space exploration methodology for higher number of operating conditions. Experimental results on industry-strength benchmarks demonstrate that: 1) compared with conventional approach OSFA could provide an average 6.1% reduction in power without performance loss; 2) NBTI-aware OSFA framework can provide significant improvement in comparison with guard-band based traditional NBTI-aware gate sizing approach; and 3) percentage savings compared to conventional methodology increases with the number of operating conditions.

Adders are the most fundamental arithmetic units, and often on the timing critical paths of microprocessors. Among various adder configurations, parallel prefix structures provide the high performance adders for higher bit-widths. With aggressive technology scaling, the performance of a parallel prefix adder, in addition to the dependence on the logic-level, is determined by wire-length and congestion which can be mitigated by adjusting fan-out. This paper proposes a polynomial-time algorithm to synthesize n bit parallel prefix adders targeting the minimization of the size of the prefix graph with log2n logic level and any arbitrary fan-out restriction. The design space exploration by our algorithm provides a set of Pareto-optimal solutions for delay vs. power trade-off, and these Pareto-optimal solutions can be used in high-performance designs instead of picking from a fixed library (Kogge Stone, Sklansky etc.). Experimental results demonstrate that our approach (i) excels highly competitive industry standard Synopsys Design Compiler adder (128 bit) in performance (2%), area (25%) and power (13.3%) in 32nm technology node, and (ii) improves performance/area over even 64 bit custom designed adders targeting 22nm technology library and implemented in an industrial high-performance design.

Modern SoCs and microprocessors, e.g., those in smart phones and laptops, typically have multiple operating conditions, such as video streaming, web browsing, standby, and so on. They will have different performance targets and run under different supply voltages. Gate sizing (with threshold voltage assignment) is a fundamental step for power/performance optimization. However, conventional gate sizing algorithms only consider one scenario, e.g., the performance-critical operating condition, which may be over-design for other operating conditions. In this paper, we present a new paradigm of gate sizing, OSFA (One-Size-Fits-All), which performs power/performance optimizations across multiple operating conditions. Based on OSFA, we also adjust the supply voltage targeting overall power optimization. Experimental results on industry-strength benchmarks demonstrate that compared with conventional approach OSFA could provide an average 6.1% reduction in power without performance loss.

With aggressive technology scaling in nanometer regime, a significant fraction of dynamic power is consumed in the clock network due to its high switching activity. Clock networks are typically synthesized and routed to optimize for zero clock skew. However, clock skew optimization is often accompanied with routing overhead which increases the clock net capacitance thereby consuming more power. In this paper, we propose a skew bounded buffer tree resynthesis algorithm to optimize clock net capacitance after the clock network has been synthesized and routed. Our algorithm restricts the skew of the designs within a specified margin from its original skew, and does not introduce any additional Design Rule Check (DRC) violation. Experimental results on industrial designs, with clock networks synthesized and routed by an industrial tool, have demonstrated that our approach can achieve an average reduction of 5.6% and 3.5% in clock net capacitance and clock dynamic power respectively with a marginal overhead in the clock skew.

Negative Bias Temperature Instability (NBTI) has become a major reliability concern in nanoscale designs. Although several previous studies have been proposed to address the NBTI effect during logic synthesis, their performance is limited because of focusing on a certain logic synthesis stage. Additionally, their complicated algorithms are not scalable to large designs. To tackle this, we propose a coordinated and scalable logic synthesis approach, which integrates techniques at different logic synthesis stages, ranging from subject graph to technology mapping and mapped netlist, to achieve an effective NBTI reduction. To our best knowledge, this is the first work that considers and mitigates NBTI impact in subject graphs, the earlier stage of logic synthesis. Experimental results on industry-strength benchmarks show that our approach can achieve 6.5% NBTI delay reduction with merely 2.5% area overhead on average, while a previous work barely gets NBTI delay reduction when the circuits are optimized beforehand, the circuit sizes are large, and standard cell libraries are richer.

With aggressive technology scaling and complex design scenarios, timing closure has become a challenging and tedious job for the designers. Timing violations persist for multi-corner, multi-mode designs in the deep-routing stage although careful optimization has been applied at every step after synthesis. Useful clock skew optimization has been suggested as an effective way to achieve design convergence and timing closure. Existing approaches on useful skew optimization: 1) calculate clock skew at sequential elements before the actual tree is synthesized and 2) do not account for the implementability of the calculated schedules at the later stages of design cycle. In this paper, we propose a novel clock tree resynthesis methodology which is based on a skew scheduling engine which works on an already built clock tree. The output of the engine is a set of positive and negative offsets which translate to the delay and accelerations, respectively in clock arrival at the clock tree pins. We demonstrate the effectiveness of the offsets at the output pins of the leaf-level clock drivers in comparison to the traditional clock scheduling in the clock pins of the flip-flops due to the better implementability and lesser area overhead and present an algorithm to accurately realize these offsets in the clock tree. Experimental results on large-scale industrial designs demonstrate that our clock tree resynthesis methodology achieves respectively 57%, 12%, and 42% average improvement in total negative slack, worst negative slack, and failure-end-point with an average overhead of 26% in clock tree area. We also experimentally study the impact of on-chip-variation-derates on our approach in terms of the timing metric improvement and clock tree overhead.

With aggressive technology scaling and complex design scenarios, timing closure has become a challenging and tedious job for the designers. Timing violations persist for multi- corner, multi-mode designs in the deep-routing stage although careful optimization has been applied at every step after synthesis. Useful clock skew optimization has been suggested as an effective way to achieve design convergence and timing closure. Existing approaches on useful skew optimization (i) calculate clock skew at sequential elements before the actual tree is synthesized, and (ii) do not account for the implementability of the calculated schedules at the later stages of design cycle. Our approach is based on a skew scheduling engine which works on an already built clock tree. The output of the engine is a set of positive and negative offsets which translate to the delay and accelerations respectively in clock arrival at the clock tree pins. A novel algorithm is presented to accurately realize these offsets in the clock tree. Experimental results on large-scale industrial designs demonstrate that our approach achieves respectively 57%, 12% and 42% average improvement in total negative slack (TNS), worst negative slack (WNS) and failure-end-point (FEP) with an average overhead of 26% in clock tree area.

This paper proposes an efficient algorithm to synthesize prefix graph structures that yield adders with the best performance-area trade-off. For designing a parallel prefix adder of a given bit-width, our approach generates prefix graph structures to optimize an objective function such as size of prefix graph subject to constraints like bit-wise output logic level. Given bit-width n and level (L) restriction, our algorithm excels the existing algorithms in minimizing the size of the prefix graph. We also prove its size-optimality when n is a power of two and L= log2n. Besides prefix graph size optimization and having the best performance-area trade-off, our approach, unlike existing techniques, can 1) handle more complex constraints such as maximum node fanout or wire-length that impact the performance/area of a design and 2) generate several feasible solutions that minimize the objective function. Generating several size-optimal solutions provides the option to choose adder designs that mitigate constraints such as wire congestion or power consumption that are difficult to model as constraints during logic synthesis. Experimental results demonstrate that our approach improves performance by 3% and area by 9% over even a 64-bit full custom designed adder implemented in an industrial high-performance design.

This paper proposes an efficient algorithm to synthesize prefix graph structures that yield adders with the best performance-area trade-off. For designing a parallel prefix adder of a given bit-width, our approach generates prefix graph structures to optimize an objective function such as size of prefix graph subject to constraints like bit-wise output logic level. Besides having the best performance-area trade-off our approach, unlike existing techniques, can (i) handle more complex constraints such as maximum node fanout or wire-length that impact the performance/area of a design and (ii) generate several feasible solutions that minimize the objective function. Generating several optimal solutions provides the option to choose adder designs that mitigate constraints such as wire congestion or power consumption that are difficult to model as constraints during logic synthesis. Experimental results demonstrate that our approach improves performance by 3% and area by 9% over even a 64-bit full custom designed adder implemented in an industrial high-performance design.