This paper presents a structure simplification procedure that allows efficient energy and accuracy tradeoffs in implementation of trained deep neural networks (DNNs). This structure simplification procedure identifies and eliminates redundant neurons in any layer of the DNN based on the output values of these neurons over the training samples. This procedure may be applied to all layers of a DNN. Pareto-optimal points in terms of accuracy and energy consumption can be easily identified from the configurations across different layers. After redundant neurons are discarded, the weights connect to the remaining neurons will be updated using matrix multiplication without retraining. Yet, retraining may still be applied if desired to further fine tune the performance. In our experiments, we show energy-accuracy tradeoff provides clear guidance to achieve efficient realization of trained DNNs. We also observe significant implementation cost reductions with up to 11.50X in energy and 12.30X in memory while the relative performance (accuracy) loss is negligible. We also discuss and validate strategies to apply our procedure to multiple layers in a DNN in order to maximize the compression ratio of the network after simplification subject to an overall budget in degradation of accuracy across all the simplified layers.
Logic synthesis is a fundamental step in hardware design whose goal is to find structural representations of Boolean functions while minimizing delay and area. If the function is completely-specified, the implementation accurately represents the function. If the function is incompletely-specified, the implementation has to be true only on the care set. While most of the algorithms in logic synthesis rely on SAT and Boolean methods to exactly implement the care set, we investigate learning in logic synthesis, attempting to trade exactness for generalization. This work is directly related to machine learning where the care set is the training set and the implementation is expected to generalize on a validation set. We present learning incompletely-specified functions based on the results of a competition conducted at IWLS 2020. The goal of the competition was to implement 100 functions given by a set of care minterms for training, while testing the implementation using a set of validation minterms sampled from the same function. We make this benchmark suite available and offer a detailed comparative analysis of the different approaches to learning.
4 October 2017 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Split manufacturing is a technique that allows manufacturing the transistor-level and lower metal layers of an integrated circuit (IC) at a high-end, untrusted foundry, while manufacturing only the higher metal layers at a smaller, trusted foundry. Using split manufacturing is only viable if the untrusted foundry cannot reverse engineer the higher metal layer connections (and thus the overall IC design) from the lower layers. This paper studies the effectiveness of proximity attack as a key step to reverse engineer a design at the untrusted foundry. We propose and study different proximity attacks based on how a set of candidates are defined for each broken connection. The attacks use both placement and routing information along with factors which capture the router’s behavior such as per-layer routing congestion. Our studies are based on designs having millions of nets routed across nine metal layers and significant layer-by-layer wire size variation. Our results show that a common, Hamming distance-based proximity attack seldom achieves a match rate over 5%. But our proposed attack yields a relatively small list of candidates which often contains the correct match. Finally, we propose a procedure to artificially insert routing blockages in a design at a desired split level, without causing any area overhead, in order to trick the router to make proximity-based reverse engineering significantly more challenging.
6 August 2019 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
This paper is the first to analyze the security of split manufacturing using machine learning (ML), based on data collected from layouts provided by industry, with eight routing metal layers and significant variation in wire size and routing congestion across the layers. Many types of layout features are considered in our ML model, including those obtained from placement, routing, and cell sizes. Since the runtime cost of our basic ML procedure becomes prohibitively large for lower layers, we propose novel techniques to make it scalable with little sacrifice in the effectiveness of the attack. Moreover, we further improve the performance in the top routing layer by making use of higher quality training samples and by exploiting the routing convention. We also propose a validation-based proximity attack procedure, which generally outperforms our recent prior work. In the experiments, we analyze the ranking of the features used in our ML model and show how features vary in importance when moving to the lower layers. We provide comprehensive evaluation and comparison of our model with different configurations and demonstrate dramatically better performance of attacks compared to the prior work.
27 June 2005 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
This paper presents a novel approach for leakage optimization through simultaneous V/sub t/ selection and assignment. V/sub t/ selection implies deciding the right value for V/sub t/ and assignment implies deciding which gates should be assigned a particular threshold voltage. We also include the effect of variability in threshold voltage on delay and leakage due to fabrication process variations in our formulations and present a scheme that lets the designer control the leakage and delay variability in his design. The proposed algorithm is a general mathematical formulation that has been shown to trivially extend to multiple threshold voltages.
Given a directed acyclic graph and different possible implementations for each node, the implementation selection problem (ISP) selects the appropriate implementation for each node such that a given global design objective is optimized, ISP is a generic formulation that is explicitly or implicitly solved in several design automation problems like leakage optimization using dual V/sub th/, gate sizing, etc. An implementation of a node results in an associated delay and perhaps cost for the node. In the presence of different sources of uncertainty and fabrication variability, fixed estimates of delays and costs of a node are extremely erroneous. We investigate a probabilistic approach to solve ISP by considering probability density functions for delays and costs of a node. We propose a dynamic-programming based approach in a probabilistic sense and introduce effective pruning criteria when dealing with probability distributions for identifying co-optimal solution at each stage. A case study of leakage optimization using dual V/sub th/ is presented where we show the effectiveness of a probabilistic approach considering V/sub th/ variability over a traditional deterministic one.