Matching Items (74)
Filtering by
- Status: Published
Description
Multidimensional (MD) discrete Fourier transform (DFT) is a key kernel algorithm in many signal processing applications, such as radar imaging and medical imaging. Traditionally, a two-dimensional (2-D) DFT is computed using Row-Column (RC) decomposition, where one-dimensional (1-D) DFTs are computed along the rows followed by 1-D DFTs along the columns. However, architectures based on RC decomposition are not efficient for large input size data which have to be stored in external memories based Synchronous Dynamic RAM (SDRAM). In this dissertation, first an efficient architecture to implement 2-D DFT for large-sized input data is proposed. This architecture achieves very high throughput by exploiting the inherent parallelism due to a novel 2-D decomposition and by utilizing the row-wise burst access pattern of the SDRAM external memory. In addition, an automatic IP generator is provided for mapping this architecture onto a reconfigurable platform of Xilinx Virtex-5 devices. For a 2048x2048 input size, the proposed architecture is 1.96 times faster than RC decomposition based implementation under the same memory constraints, and also outperforms other existing implementations. While the proposed 2-D DFT IP can achieve high performance, its output is bit-reversed. For systems where the output is required to be in natural order, use of this DFT IP would result in timing overhead. To solve this problem, a new bandwidth-efficient MD DFT IP that is transpose-free and produces outputs in natural order is proposed. It is based on a novel decomposition algorithm that takes into account the output order, FPGA resources, and the characteristics of off-chip memory access. An IP generator is designed and integrated into an in-house FPGA development platform, AlgoFLEX, for easy verification and fast integration. The corresponding 2-D and 3-D DFT architectures are ported onto the BEE3 board and their performance measured and analyzed. The results shows that the architecture can maintain the maximum memory bandwidth throughout the whole procedure while avoiding matrix transpose operations used in most other MD DFT implementations. The proposed architecture has also been ported onto the Xilinx ML605 board. When clocked at 100 MHz, 2048x2048 images with complex single-precision can be processed in less than 27 ms. Finally, transpose-free imaging flows for range-Doppler algorithm (RDA) and chirp-scaling algorithm (CSA) in SAR imaging are proposed. The corresponding implementations take advantage of the memory access patterns designed for the MD DFT IP and have superior timing performance. The RDA and CSA flows are mapped onto a unified architecture which is implemented on an FPGA platform. When clocked at 100MHz, the RDA and CSA computations with data size 4096x4096 can be completed in 323ms and 162ms, respectively. This implementation outperforms existing SAR image accelerators based on FPGA and GPU.
ContributorsYu, Chi-Li (Author) / Chakrabarti, Chaitali (Thesis advisor) / Papandreou-Suppappola, Antonia (Committee member) / Karam, Lina (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2012
Description
Today's mobile devices have to support computation-intensive multimedia applications with a limited energy budget. In this dissertation, we present architecture level and algorithm-level techniques that reduce energy consumption of these devices with minimal impact on system quality. First, we present novel techniques to mitigate the effects of SRAM memory failures in JPEG2000 implementations operating in scaled voltages. We investigate error control coding schemes and propose an unequal error protection scheme tailored for JPEG2000 that reduces overhead without affecting the performance. Furthermore, we propose algorithm-specific techniques for error compensation that exploit the fact that in JPEG2000 the discrete wavelet transform outputs have larger values for low frequency subband coefficients and smaller values for high frequency subband coefficients. Next, we present use of voltage overscaling to reduce the data-path power consumption of JPEG codecs. We propose an algorithm-specific technique which exploits the characteristics of the quantized coefficients after zig-zag scan to mitigate errors introduced by aggressive voltage scaling. Third, we investigate the effect of reducing dynamic range for datapath energy reduction. We analyze the effect of truncation error and propose a scheme that estimates the mean value of the truncation error during the pre-computation stage and compensates for this error. Such a scheme is very effective for reducing the noise power in applications that are dominated by additions and multiplications such as FIR filter and transform computation. We also present a novel sum of absolute difference (SAD) scheme that is based on most significant bit truncation. The proposed scheme exploits the fact that most of the absolute difference (AD) calculations result in small values, and most of the large AD values do not contribute to the SAD values of the blocks that are selected. Such a scheme is highly effective in reducing the energy consumption of motion estimation and intra-prediction kernels in video codecs. Finally, we present several hybrid energy-saving techniques based on combination of voltage scaling, computation reduction and dynamic range reduction that further reduce the energy consumption while keeping the performance degradation very low. For instance, a combination of computation reduction and dynamic range reduction for Discrete Cosine Transform shows on average, 33% to 46% reduction in energy consumption while incurring only 0.5dB to 1.5dB loss in PSNR.
ContributorsEmre, Yunus (Author) / Chakrabarti, Chaitali (Thesis advisor) / Bakkaloglu, Bertan (Committee member) / Cao, Yu (Committee member) / Papandreou-Suppappola, Antonia (Committee member) / Arizona State University (Publisher)
Created2012
Description
Circuits on smaller technology nodes become more vulnerable to radiation-induced upset. Since this is a major problem for electronic circuits used in space applications, designers have a variety of solutions in hand. Radiation hardening by design (RHBD) is an approach, where electronic components are designed to work properly in certain radiation environments without the use of special fabrication processes. This work focuses on the cache design for a high performance microprocessor. The design tries to mitigate radiation effects like SEE, on a commercial foundry 45 nm SOI process. The design has been ported from a previously done cache design at the 90 nm process node. The cache design is a 16 KB, 4 way set associative, write-through design that uses a no-write allocate policy. The cache has been tested to write and read at above 2 GHz at VDD = 0.9 V. Interleaved layout, parity protection, dual redundancy, and checking circuits are used in the design to achieve radiation hardness. High speed is accomplished through the use of dynamic circuits and short wiring routes wherever possible. Gated clocks and optimized wire connections are used to reduce power. Structured methodology is used to build up the entire cache.
ContributorsXavier, Jerin (Author) / Clark, Lawrence T (Thesis advisor) / Cao, Yu (Committee member) / Allee, David R. (Committee member) / Arizona State University (Publisher)
Created2012
Description
Negative bias temperature instability (NBTI) is a leading aging mechanism in modern digital and analog circuits. Recent NBTI data exhibits an excessive amount of randomness and fast recovery, which are difficult to be handled by conventional power-law model (tn). Such discrepancies further pose the challenge on long-term reliability prediction under statistical variations and Dynamic Voltage Scaling (DVS) in real circuit operation. To overcome these barriers, the modeling effort in this work (1) practically explains the aging statistics due to randomness in number of traps with log(t) model, accurately predicting the mean and variance shift; (2) proposes cycle-to-cycle model (from the first-principle of trapping) to handle aging under multiple supply voltages, predicting the non-monotonic behavior under DVS (3) presents a long-term model to estimate a tight upper bound of dynamic aging over multiple cycles, and (4) comprehensively validates the new set of aging models with 65nm statistical silicon data. Compared to previous models, the new set of aging models capture the aging variability and the essential role of the recovery phase under DVS, reducing unnecessary guard-banding during the design stage. With CMOS technology scaling, design for reliability has become an important step in the design cycle, and increased the need for efficient and accurate aging simulation methods during the design stage. NBTI induced delay shifts in logic paths are asymmetric in nature, as opposed to averaging effect due to recovery assumed in traditional aging analysis. Timing violations due to aging, in particular, are very sensitive to the standby operation regime of a digital circuit. In this report, by identifying the critical moments in circuit operation and considering the asymmetric aging effects, timing violations under NBTI effect are correctly predicted. The unique contributions of the simulation flow include: (1) accurate modeling of aging induced delay shift due to threshold voltage (Vth) shift using only the delay dependence on supply voltage from cell library; (2) simulation flow for asymmetric aging analysis is proposed and conducted at critical points in circuit operation; (3) setup and hold timing violations due to NBTI aging in logic and clock buffer are investigated in sequential circuits and (4) proposed framework is tested in VLSI applications such DDR memory circuits. This methodology is comprehensively demonstrated with ISCAS89 benchmark circuits using a 45nm Nangate standard cell library characterized using predictive technology models. Our proposed design margin assessment provides design insights and enables resilient techniques for mitigating digital circuit aging.
ContributorsVelamala, Jyothi Bhaskarr (Author) / Cao, Yu (Thesis advisor) / Clark, Lawrence (Committee member) / Chakrabarti, Chaitali (Committee member) / Yu, Hongbin (Committee member) / Arizona State University (Publisher)
Created2012
Description
RRAM is an emerging technology that looks to replace FLASH NOR and possibly NAND memory. It is attractive because it uses an adjustable resistance and does not rely on charge; in the sub-10nm feature size circuitry this is critical. However, RRAM cross-point arrays suffer tremendously from leakage currents that prevent proper readings in larger array sizes. In this research an exponential IV selector was added to each cell to minimize this current. Using this technique the largest array-size supportable was determined to be 512x512 cells using the conventional voltage sense amplifier by HSPICE simulations. However, with the increase in array size, the sensing latency also remarkably increases due to more sneak path currents, approaching 873 ns for the 512x512 array.
ContributorsMadler, Ryan Anton (Author) / Yu, Shimeng (Thesis director) / Cao, Yu (Committee member) / Electrical Engineering Program (Contributor) / Barrett, The Honors College (Contributor)
Created2016-05
Description
In this thesis two methodologies have been proposed for evaluating the fault response of analog/RF circuits. These proposed approaches are used to evaluate the response of the faulty circuit in terms of specifications/measurements. Faulty response can be used to evaluate important test metrics like fail probability, fault coverage and yield coverage of given measurements under process variations. Once the models for faulty and fault free circuit are generated, one needs to perform Monte Carlo sampling (as opposed to Monte Carlo simulations) to compute these statistical parameters with high accuracy. The first method is based on adaptively determining the order of the model based on the error budget in terms of computing the statistical metrics and position of the threshold(s) to decide how precisely necessary models need to be extracted. In the second method, using hierarchy in process variations a hybrid of heuristics and localized linear models have been proposed. Experiments on LNA and Mixer using the adaptive model order selection procedure can reduce the number of necessary simulations by 7.54x and 7.03x respectively in the computation of fail probability for an error budget of 2%. Experiments on LNA using the hybrid approach can reduce the number of necessary simulations by 21.9x and 17x for four and six output parameters cases for improved accuracy in test statistics estimation.
ContributorsSubrahmaniyan Radhakrishnan, Gurusubrahmaniyan (Author) / Ozev, Sule (Thesis advisor) / Blain Christen, Jennifer (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2010
Description
To extend the lifetime of complementary metal-oxide-semiconductors (CMOS), emerging process techniques are being proposed to conquer the manufacturing difficulties. New structures and materials are proposed with superior electrical properties to traditional CMOS, such as strain technology and feedback field-effect transistor (FB-FET). To continue the design success and make an impact on leading products, advanced circuit design exploration must begin concurrently with early silicon development. Therefore, an accurate and scalable model is desired to correctly capture those effects and flexible to extend to alternative process choices. For example, strain technology has been successfully integrated into CMOS fabrication to improve transistor performance but the stress is non-uniformly distributed in the channel, leading to systematic performance variations. In this dissertation, a new layout-dependent stress model is proposed as a function of layout, temperature, and other device parameters. Furthermore, a method of layout decomposition is developed to partition the layout into a set of simple patterns for model extraction. These solutions significantly reduce the complexity in stress modeling and simulation. On the other hand, semiconductor devices with self-feedback mechanisms are emerging as promising alternatives to CMOS. Fe-FET was proposed to improve the switching by integrating a ferroelectric material as gate insulator in a MOSFET structure. Under particular circumstances, ferroelectric capacitance is effectively negative, due to the negative slope of its polarization-electrical field curve. This property makes the ferroelectric layer a voltage amplifier to boost surface potential, achieving fast transition. A new threshold voltage model for Fe-FET is developed, and is further revealed that the impact of random dopant fluctuation (RDF) can be suppressed. Furthermore, through silicon via (TSV), a key technology that enables the 3D integration of chips, is studied. TSV structure is usually a cylindrical metal-oxide-semiconductors (MOS) capacitor. A piecewise capacitance model is proposed for 3D interconnect simulation. Due to the mismatch in coefficients of thermal expansion (CTE) among materials, thermal stress is observed in TSV process and impacts neighboring devices. The stress impact is investigated to support the interaction between silicon process and IC design at the early stage.
ContributorsWang, Chi-Chao (Author) / Cao, Yu (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Clark, Lawrence (Committee member) / Schroder, Dieter (Committee member) / Arizona State University (Publisher)
Created2011
Description
Due to high DRAM access latency and energy, several convolutional neural network(CNN) accelerators face performance and energy efficiency challenges, which are critical for embedded implementations. As these applications exploit larger datasets, memory accesses of these emerging applications are increasing. As a result, it is difficult to predict the combined dynamic random access memory (DRAM) workload behavior, which can sabotage memory optimizations in software. To understand the impact of external memory access on CNN accelerators which reduces the high DRAMaccess latency and energy, simulators such as RAMULATOR and VAMPIRE have been proposed in prior work. In this work, we utilize these simulators to benchmark external memory access in CNN accelerators. Experiments are performed generating trace files based on the number of parameters and data precision and also using trace file generated for CNN Accelerator Altera Arria 10 GX 1150 FPGA data to complete the end to end workflow using the mentioned simulators. Besides that, certain modifications were made in the default VAMPIRE code to implement certain functionalities such as PREA(Precharge All) and REF(Refresh). Then, precalculated energies were computed for DDR3, DDR4, and HBM based on the micron model to mention it in the dram specification file inputted to the VAMPIRE tool. An experimental study was performed and a comparison is made between DDR3, DDR4, and HBM, it was proved that DDR4 is nearly 31% more energy-efficient than DDR3 and HBMis 54% energy-efficient than DDR3. Performed modeling and experimental analysis on a large set of data and then split it into a set of data and compared the results of the small sets multiplied with the number of sets and the large data set and concluded that the results were nearly the same. Finally, a GUI is developed by wrapping both the simulators. GUI provides user-friendly access and can analyze the parameters without much prior knowledge and understanding of the working.
ContributorsPannala, Manvitha (Author) / Cao, Yu (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Seo, Jae-Sun (Committee member) / Arizona State University (Publisher)
Created2021
Description
Convolutional neural networks(CNNs) achieve high accuracy on large datasets but requires significant computation and storage requirement for training/testing. While many applications demand low latency and energy-efficient processing of the images, deploying these complex algorithms on the hardware is a challenging task. This dissertation first presents a compiler-based CNN training accelerator using DDR3 and HBM2 memory. An optimized RTL library is implemented to perform training-specific tasks and an RTL compiler is developed to generate FPGA-synthesizable RTL based on user-defined constraints. High Bandwidth Memory(HBM) provides efficient off-chip communication and improves the training performance. The impact of HBM2 on CNN training workloads is analyzed and compressively compared with DDR3. For training ResNet-20/VGG-like CNNs for the CIFAR-10 dataset, the proposed CNN training accelerator on Stratix-10 GX FPGA(DDR3) demonstrates 479 GOPS performance, and on Stratix-10 MX FPGA(HBM) shows 4.5/9.7 X energy-efficiency improvement compared to Tesla V100 GPU. Next, the FPGA online learning accelerator is presented. Adopting model segmentation techniques from Progressive Segmented Training(PST), the online learning accelerator achieved a 4.2X reduction in training latency. Furthermore, this dissertation presents an 8-bit floating-point (FP8) training processor which implements (1) Highly parallel tensor cores that maintain high PE utilization, (2) Hardware-efficient channel gating for dynamic output activation sparsity (3) Dynamic weight sparsity based on group Lasso (4) Gradient skipping based on FP prediction error. The 28nm prototype chip demonstrates significant improvements in FLOPs reduction (7.3×), energy efficiency (16.4 TFLOPS/W), and overall training latency speedup (4.7×) for both supervised training and self-supervised training tasks. In addition to the training accelerators, this dissertation also presents a CNN inference accelerator on ASIC(FixyNN) and FPGA(FixyFPGA). FixyNN consists of a fixed-weight feature extractor that generates ubiquitous CNN features and a conventional programmable CNN accelerator. In the fixed-weight feature extractor, the network weights are hard-coded into hardware and used as a fixed operand for the multiplication. Experimental results demonstrate FixyNN can achieve very high energy efficiencies up to 26.6 TOPS/W, and FixyFPGA achieves $2.34\times$ higher GOPS on ImageNet classification. In summary, this dissertation comprehensively discusses novel architectures of high-performance and energy-efficient ASIC/FPGA CNN inference/training accelerators.
ContributorsKolala Venkataramaniah, Shreyas (Author) / Seo, Jae-Sun (Thesis advisor) / Cao, Yu (Committee member) / Chakrabarti, Chaitali (Committee member) / Fan, Deliang (Committee member) / Arizona State University (Publisher)
Created2022
Description
Adversarial threats of deep learning are increasingly becoming a concern due to the ubiquitous deployment of deep neural networks(DNNs) in many security-sensitive domains. Among the existing threats, adversarial weight perturbation is an emerging class of threats that attempts to perturb the weight parameters of DNNs to breach security and privacy.In this thesis, the first weight perturbation attack introduced is called Bit-Flip Attack (BFA), which can maliciously flip a small number of bits within a computer’s main memory system storing the DNN weight parameter to achieve malicious objectives. Our developed algorithm can achieve three specific attack objectives: I) Un-targeted accuracy degradation attack, ii) Targeted attack, & iii) Trojan attack. Moreover, BFA utilizes the rowhammer technique to demonstrate the bit-flip attack in an actual computer prototype. While the bit-flip attack is conducted in a white-box setting, the subsequent contribution of this thesis is to develop another novel weight perturbation attack in a black-box setting. Consequently, this thesis discusses a new study of DNN model vulnerabilities in a multi-tenant Field Programmable Gate Array (FPGA) cloud under a strict black-box framework. This newly developed attack framework injects faults in the malicious tenant by duplicating specific DNN weight packages during data transmission between off-chip memory and on-chip buffer of a victim FPGA. The proposed attack is also experimentally validated in a multi-tenant cloud FPGA prototype. In the final part, the focus shifts toward deep learning model privacy, popularly known as model extraction, that can steal partial DNN weight parameters remotely with the aid of a memory side-channel attack. In addition, a novel training algorithm is designed to utilize the partially leaked DNN weight bit information, making the model extraction attack more effective. The algorithm effectively leverages the partial leaked bit information and generates a substitute prototype of the victim model with almost identical performance to the victim.
ContributorsRakin, Adnan Siraj (Author) / Fan, Deliang (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Seo, Jae-Sun (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2022