In this paper, we consider Switched-Delay-Lines (SDL) constructions of optical priority queues by using optical (bufferless) crossbar switches and optical fiber delay lines. In a priority queue, each packet is associated with a priority upon its arrival, the highest-priority packet is sent out from the queue whenever there is a departure request, and the lowest-priority packet is dropped from the queue whenever there is a buffer overflow. Given any system for SDL constructions of optical priority queues, the main research problem is twofold: (i) the design of the routing policy performed by the optical crossbar switches; (ii) the choice of the delays of the optical fiber delay lines. Sarwate and Anantharam are the first to propose a feedback system consisting of an optical$(M+2)\times (M+2)$crossbar switch and M optical fiber delay lines (seeFigure 1inSection I) for SDL constructions of optical priority queues, and they have shown that the largest buffer size that can possibly be achieved by using such a feedback system is$2^{M}$. However, whether this theoretical buffer size$2^{M}$can be achieved or not remains an open research problem. Currently, the best result in the literature was obtained by Cheng et al. and the achieved buffer size is$2^{O(\sqrt {\alpha M})}$, where$\alpha $is a constant that depends on the parameters used in their constructions. In this paper, we consider a discrete-time setting and use a feedback system consisting of an optical crossbar switch and multiple groups of optical first-in first-out (FIFO) multiplexers with delay one (FM1’s) for SDL constructions of optical priority queues under apriority-based routing policy (seeFigure 2inSection I). Our contributions are as follows: (i) We extend and generalize an important class of constructions that contains the optimal constructions in the work of Cheng et al. As a result, we achieve larger buffer sizes and less construction complexities/costs than those by Cheng et al. (ii) We obtain a closed-form expression for the maximum buffer size that is achieved by the optimal construction for the scenario that each group of FM1’s has the same number of FM1’s. (iii) Our constructions possess a salient feature, namely, fault-tolerant capability, that can tolerate the malfunctioning of some FM1’s by using the generalized results obtained in this paper. (iv) We show that our constructions can be implemented by using an optical$(M+2)\times (M+2)$crossbar switch andMoptical fiber delay lines, and achieve a buffer size$2^{O(\sqrt {\alpha M})}$, where$\alpha $is a constant that depends on the parameters used in our constructions and is better, i.e., larger, than that in the work of Cheng et al. in a very broad regime.

GDBN (good die in bad neighborhood) methodology has been regarded as an effective technique for reducing DPPM (defect parts per million), by identifying and rejecting suspicious dice even though they test good. Instead of examining eight immediate neighbors or exploiting simple linear regression, in this paper we propose to employ a window of larger size for broad-sighted recognition of neighborhood, and make best use of the larger window for accurate prediction of the suspicious level for any given die. The proposed methodology is realized by using an artificial neural network (NN), and is a breakthrough of NN-based work for solving the problem of GDBN. Various experiments on two sets of data clearly reveal the superiority of our NN-based methodology over other existing methods. Besides reducing DPPM, our methodology is able to achieve 1. 5X-2X better reduction in the cost for return merchandise authorization (RMA).

In order to reduce DPPM (defect parts per million), IDDQ testing methodology can be exploited for identifying "outliers" which are potentially defective but not detected by signoff functional and parametric tests. Conventional IDDQ testing paradigms depending on a simple statistical 6σ rule or engineers’ experience are usually too conservative to effectively identify non-trivial outliers, especially when spatial correlations are of great concern/influence. In this paper, by employing a stochastic regression model, the mean as well as the variance of the IDDQ of a die under test (DUT) can be predicted. According to the predicted mean and variance, we derive an expected IDDQ range and identify the DUT as an outlier if its actual IDDQ measurement is beyond the expected range. The proposed stochastic regression model is obtained by training a convolutional neural network (CNN) and, based on its primitive property of convolutional kernel mapping with large volume of industrial data, spatial correlations (due to spatially-correlated process variations, etc) can be considered/captured. The trained data-driven CNN is highly accurate in terms of R-square (0.958) and RMSE (0.783), and the percentage of identified outliers (0.047%) is very close to the theoretical reference (0.050%), which validates the efficacy of our proposed methodology.