@misc{WenBaroniPerezetal., author = {Wen, Jianan and Baroni, Andrea and Perez, Eduardo and Uhlmann, Max and Fritscher, Markus and KrishneGowda, Karthik and Ulbricht, Markus and Wenger, Christian and Krstic, Milos}, title = {Towards reliable and energy-efficient RRAM based discrete fourier transform accelerator}, series = {2024 Design, Automation \& Test in Europe Conference \& Exhibition (DATE)}, journal = {2024 Design, Automation \& Test in Europe Conference \& Exhibition (DATE)}, publisher = {IEEE}, isbn = {978-3-9819263-8-5}, issn = {1558-1101}, doi = {10.23919/DATE58400.2024.10546709}, pages = {1 -- 6}, abstract = {The Discrete Fourier Transform (DFT) holds a prominent place in the field of signal processing. The development of DFT accelerators in edge devices requires high energy efficiency due to the limited battery capacity. In this context, emerging devices such as resistive RAM (RRAM) provide a promising solution. They enable the design of high-density crossbar arrays and facilitate massively parallel and in situ computations within memory. However, the reliability and performance of the RRAM-based systems are compromised by the device non-idealities, especially when executing DFT computations that demand high precision. In this paper, we propose a novel adaptive variability-aware crossbar mapping scheme to address the computational errors caused by the device variability. To quantitatively assess the impact of variability in a communication scenario, we implemented an end-to-end simulation framework integrating the modulation and demodulation schemes. When combining the presented mapping scheme with an optimized architecture to compute DFT and inverse DFT(IDFT), compared to the state-of-the-art architecture, our simulation results demonstrate energy and area savings of up to 57 \% and 18 \%, respectively. Meanwhile, the DFT matrix mapping error is reduced by 83\% compared to conventional mapping. In a case study involving 16-quadrature amplitude modulation (QAM), with the optimized architecture prioritizing energy efficiency, we observed a bit error rate (BER) reduction from 1.6e-2 to 7.3e-5. As for the conventional architecture, the BER is optimized from 2.9e-3 to zero.}, language = {en} } @misc{UhlmannPerezBoschQuesadaFritscheretal., author = {Uhlmann, Max and P{\´e}rez-Bosch Quesada, Emilio and Fritscher, Markus and P{\´e}rez, Eduardo and Schubert, Markus Andreas and Reichenbach, Marc and Ostrovskyy, Philip and Wenger, Christian and Kahmen, Gerhard}, title = {One-Transistor-Multiple-RRAM Cells for Energy-Efficient In-Memory Computing}, series = {21st IEEE Interregional NEWCAS Conference (NEWCAS)}, journal = {21st IEEE Interregional NEWCAS Conference (NEWCAS)}, publisher = {Institute of Electrical and Electronics Engineers (IEEE)}, isbn = {979-8-3503-0024-6}, issn = {2474-9672}, doi = {10.1109/NEWCAS57931.2023.10198073}, pages = {5}, abstract = {The use of resistive random-access memory (RRAM) for in-memory computing (IMC) architectures has significantly improved the energy-efficiency of artificial neural networks (ANN) over the past years. Current RRAM-technologies are physically limited to a defined unambiguously distinguishable number of stable states and a maximum resistive value and are compatible with present complementary metal-oxide semiconductor (CMOS)-technologies. In this work, we improved the accuracy of current ANN models by using increased weight resolutions of memristive devices, combining two or more in-series RRAM cells, integrated in the back end of line (BEOL) of the CMOS process. Based on system level simulations, 1T2R devices were fabricated in IHP's 130nm SiGe:BiCMOS technology node, demonstrating an increased number of states. We achieved an increase in weight resolution from 3 bit in ITIR cells to 6.5 bit in our 1T2R cell. The experimental data of 1T2R devices gives indications for the performance and energy-efficiency improvement in ITNR arrays for ANN applications.}, language = {en} } @misc{FritscherKnoedtelMallahetal., author = {Fritscher, Markus and Kn{\"o}dtel, Johannes and Mallah, Maen and Pechmann, Stefan and Perez-Bosch Quesada, Emilio and Rizzi, Tommaso and Wenger, Christian and Reichenbach, Marc}, title = {Mitigating the Effects of RRAM Process Variation on the Accuracy of Artifical Neural Networks}, series = {Embedded Computer Systems: Architectures, Modeling, and Simulation. SAMOS 2021. Lecture Notes in Computer Science}, journal = {Embedded Computer Systems: Architectures, Modeling, and Simulation. SAMOS 2021. Lecture Notes in Computer Science}, publisher = {Springer}, isbn = {978-3-031-04579-0}, issn = {0302-9743}, doi = {10.1007/978-3-031-04580-6_27}, pages = {401 -- 417}, abstract = {Weight storage is a key challenge in the efficient implementation of artificial neural networks. Novel memory technologies such as RRAM are able to greatly improve density and introduce non-volatility and multibit capabilities to this component of ANN accelerators. The usage of RRAM in this domain comes with downsides, mainly caused by cycle-to-cycle and device-to-device variability leading to erroneous readouts, greatly affecting digital systems. ANNs have the ability to compensate for this by their inherent redundancy and usually exhibit a gradual deterioration in the accuracy of the task at hand. This means, that slight error rates can be acceptable for weight storage in an ANN accelerator. In this work we link device-to-device variability to the accuracy of an ANN for such an accelerator. From this study, we can estimate how strongly a certain net is affected by a certain device parameter variability. This methodology is then used to present three mitigation strategies and to evaluate how they affect the reaction of the network to variability: a) Dropout Layers b) Fault-Aware Training c) Redundancy. These mitigations are then evaluated by their ability to improve accuracy and to lower hardware overhead by providing data for a real-word example. We improved this network's resilience in such a way that it could tolerate double the variation in one of the device parameters (standard deviation of the oxide thickness can be 0.4 nm instead of 0.2 nm while maintaining sufficient accuracy.)}, language = {en} } @inproceedings{WenVargasZhuetal., author = {Wen, Jianan and Vargas, Fabian Luis and Zhu, Fukun and Reiser, Daniel and Baroni, Andrea and Fritscher, Markus and P{\´e}rez, Eduardo and Reichenbach, Marc and Wenger, Christian and Krstic, Milos}, title = {Cycle-Accurate FPGA Emulation of RRAM Crossbar Array: Efficient Device and Variability Modeling with Energy Consumption Assessment}, series = {2024 IEEE 25th Latin American Test Symposium (LATS)}, booktitle = {2024 IEEE 25th Latin American Test Symposium (LATS)}, publisher = {IEEE}, doi = {10.1109/LATS62223.2024.10534601}, pages = {6}, abstract = {Emerging device technologies such as resistive RAM (RRAM) are increasingly recognized in enhancing system performance, particularly in applications demanding extensive vector-matrix multiplications (VMMs) with high parallelism. However, a significant limitation in current electronics design automation (EDA) tools is their lack of support for rapid prototyping, design space exploration, and the integration of inherent process-dependent device variability into system-level simulations, which is essential for assessing system reliability. To address this gap, we introduce a field-programmable gate array (FPGA) based emulation approach for RRAM crossbars featuring cycle-accurate emulations in real time without relying on complex device models. Our approach is based on pre-generated look-up tables (LUTs) to accurately represent the RRAM device behavior. To efficiently model the device variability at the system level, we propose using the multivariate kernel density estimation (KDE) method to augment the measured RRAM data. The proposed emulator allows precise latency determination for matrix mapping and computation operations. Meanwhile, by coupling with the NeuroSim framework, the corresponding energy consumption can be estimated. In addition to facilitating a range of in-depth system assessments, experimental results suggest a remarkable reduction of emulation time compared to the classic behavioral simulation.}, language = {en} } @misc{JiaPechmannMarkusetal., author = {Jia, Ruolan and Pechmann, Stefan and Markus, Fritscher and Wenger, Christian and Zhang, Lei and Hagelauer, Amelie}, title = {Soft-Error Analysis of RRAM 1T1R Compute-In-Memory Core for Artificial Neural Networks}, series = {2024 39th Conference on Design of Circuits and Integrated Systems (DCIS)}, journal = {2024 39th Conference on Design of Circuits and Integrated Systems (DCIS)}, publisher = {IEEE}, doi = {10.1109/DCIS62603.2024.10769203}, pages = {1 -- 5}, abstract = {This work analyses SEU-induced soft-errors in analog compute-in-memory cores using resistive random-access memory (RRAM) for artificial neural networks, where their bitcells utilize one-transistor-one-RRAM (1T1R) structure. This is modeled by combining the Stanford-PKU RRAM Model and the model of the radiation-induced photocurrent in access transistors. As results, this work derives the maximal RRAM crossbar size without occurring any logic flip and indicates the requirements for RRAM technology to achieve a SEU-resilient 1T1R compute-in memory cores.}, language = {en} } @misc{FritscherSinghRizzietal., author = {Fritscher, Markus and Singh, Simranjeet and Rizzi, Tommaso and Baroni, Andrea and Reiser, Daniel and Mallah, Maen and Hartmann, David and Bende, Ankit and Kempen, Tim and Uhlmann, Max and Kahmen, Gerhard and Fey, Dietmar and Rana, Vikas and Menzel, Stephan and Reichenbach, Marc and Krstic, Milos and Merchant, Farhad and Wenger, Christian}, title = {A flexible and fast digital twin for RRAM systems applied for training resilient neural networks}, series = {Scientific Reports}, volume = {14}, journal = {Scientific Reports}, number = {1}, publisher = {Springer Science and Business Media LLC}, issn = {2045-2322}, doi = {10.1038/s41598-024-73439-z}, pages = {13}, abstract = {Resistive Random Access Memory (RRAM) has gained considerable momentum due to its non-volatility and energy efficiency. Material and device scientists have been proposing novel material stacks that can mimic the "ideal memristor" which can deliver performance, energy efficiency, reliability and accuracy. However, designing RRAM-based systems is challenging. Engineering a new material stack, designing a device, and experimenting takes significant time for material and device researchers. Furthermore, the acceptability of the device is ultimately decided at the system level. We see a gap here where there is a need for facilitating material and device researchers with a "push button" modeling framework that allows to evaluate the efficacy of the device at system level during early device design stages. Speed, accuracy, and adaptability are the fundamental requirements of this modelling framework. In this paper, we propose a digital twin (DT)-like modeling framework that automatically creates RRAM device models from device measurement data. Furthermore, the model incorporates the peripheral circuit to ensure accurate energy and performance evaluations. We demonstrate the DT generation and DT usage for multiple RRAM technologies and applications and illustrate the achieved performance of our GPU implementation. We conclude with the application of our modeling approach to measurement data from two distinct fabricated devices, validating its effectiveness in a neural network processing an Electrocardiogram (ECG) dataset and incorporating Fault Aware Training (FAT).}, language = {en} } @misc{FritscherWengerKrstic, author = {Fritscher, Markus and Wenger, Christian and Krstic, Milos}, title = {From device to application - integrating RRAM Accelerator Blocks into large AI systems}, series = {2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)}, journal = {2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)}, publisher = {IEEE}, isbn = {979-8-3503-5411-9}, issn = {2159-3477}, doi = {10.1109/ISVLSI61997.2024.00111}, pages = {592 -- 592}, abstract = {This work provides an introduction to design methodologies for RRAM-based systems. We illustrate the impact of device variation on the performance of neural networks and propose a circuit-level integration approach for RRAM-based compute blocks. Moreover, we demonstrate a possible architectural integration by incorporating RRAM-based VMM blocks fabricated in a 130 nm CMOS process into a RISC-V.}, language = {en} } @misc{FritscherUhlmannOstrovskyyetal., author = {Fritscher, Markus and Uhlmann, Max and Ostrovskyy, Philip and Reiser, Daniel and Chen, Junchao and Schubert, Andreas and Schulze, Carsten and Kahmen, Gerhard and Fey, Dietmar and Reichenbach, Marc and Krstic, Milos and Wenger, Christian}, title = {Area-efficient digital design using RRAM-CMOS standard cells}, series = {2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)}, volume = {18}, journal = {2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)}, publisher = {IEEE}, isbn = {979-8-3503-5411-9}, issn = {2159-3477}, doi = {10.1109/ISVLSI61997.2024.00026}, pages = {81 -- 87}, abstract = {Extending the scalability of digital integrated circuits through novel device concepts is an attractive option. Among these concepts, resistive random access memory (RRAM) devices allow fast and nonvolatile operation. However, building large memristive systems is still challenging since large analog circuits have to be designed and integrated. In this paper, we propose a novel solution - the implementation of digital standard cells by the means of RRAM devices. While this methodology is universal, with applications ranging from few-device-circuits to large macroblocks, we demonstrate it for a 2T2R-cell. The benefits of using RRAM devices are demonstrated by implementing a NAND standard cell merely consuming the area of two transistors. This cell is about 25 \% smaller than the equivalent CMOS NAND in the same technology. We use these cells to implement a half adder, beating the area of the equivalent CMOS implementation using more sophisticates gates by 15 \%. Lastly, we fully integrate this novel standard cell into a digital standard cell library and perform a synthesis and layout of a RISC-V CPU core.}, language = {en} } @misc{VishwakarmaFritscherHagelaueretal., author = {Vishwakarma, Abhinav and Fritscher, Markus and Hagelauer, Amelie and Reichenbach, Marc}, title = {An RRAM-based building block for reprogrammable non-uniform sampling ADCs}, series = {Information Technology : it}, volume = {65}, journal = {Information Technology : it}, number = {1-2}, issn = {2196-7032}, doi = {10.1515/itit-2023-0021}, pages = {39 -- 51}, language = {en} } @misc{ReiserReichenbachRizzietal., author = {Reiser, Daniel and Reichenbach, Marc and Rizzi, Tommaso and Baroni, Andrea and Fritscher, Markus and Wenger, Christian and Zambelli, Cristian and Bertozzi, Davide}, title = {Technology-Aware Drift Resilience Analysis of RRAM Crossbar Array Configurations}, series = {21st IEEE Interregional NEWCAS Conference (NEWCAS), 26-28 June 2023, Edinburgh, United Kingdom}, journal = {21st IEEE Interregional NEWCAS Conference (NEWCAS), 26-28 June 2023, Edinburgh, United Kingdom}, publisher = {IEEE}, address = {Piscataway, NJ}, isbn = {979-8-3503-0024-6}, doi = {10.1109/NEWCAS57931.2023}, abstract = {In-memory computing with resistive-switching random access memory (RRAM) crossbar arrays has the potential to overcome the major bottlenecks faced by digital hardware for data-heavy workloads such as deep learning. However, RRAM devices are subject to several non-idealities that result in significant inference accuracy drops compared with software baseline accuracy. A critical one is related to the drift of the conductance states appearing immediately at the end of program and verify algorithms that are mandatory for accurate multi-level conductance operation. The support of drift models in state-of-the-art simulation tools of memristive computationin-memory is currently only in the early stage, since they overlook key device- and array-level parameters affecting drift resilience such as the programming algorithm of RRAM cells, the choice of target conductance states and the weight-toconductance mapping scheme. The goal of this paper is to fully expose these parameters to RRAM crossbar designers as a multi-dimensional optimization space of drift resilience. For this purpose, a simulation framework is developed, which comes with the suitable abstractions to propagate the effects of those RRAM crossbar configuration parameters to their ultimate implications over inference performance stability.}, language = {en} } @misc{WenVargasZhuetal., author = {Wen, Jianan and Vargas, Fabian Luis and Zhu, Fukun and Reiser, Daniel and Baroni, Andrea and Fritscher, Markus and Perez, Eduardo and Reichenbach, Marc and Wenger, Christian and Krstic, Milos}, title = {RRAMulator : an efficient FPGA-based emulator for RRAM crossbar with device variability and energy consumption evaluation}, series = {Microelectronics Reliability}, volume = {168}, journal = {Microelectronics Reliability}, publisher = {Elsevier BV}, address = {Amsterdam}, issn = {0026-2714}, doi = {10.1016/j.microrel.2025.115630}, pages = {1 -- 10}, abstract = {The in-memory computing (IMC) systems based on emerging technologies have gained significant attention due to their potential to enhance performance and energy efficiency by minimizing data movement between memory and processing unit, which is especially beneficial for data-intensive applications. Designing and evaluating systems utilizing emerging memory technologies, such as resistive RAM (RRAM), poses considerable challenges due to the limited support from electronics design automation (EDA) tools for rapid development and design space exploration. Additionally, incorporating technology-dependent variability into system-level simulations is critical to accurately assess the impact on system reliability and performance. To bridge this gap, we propose RRAMulator, a field-programmable gate array (FPGA) based hardware emulator for RRAM crossbar array. To avoid the complex device models capturing the nonlinear current-voltage (IV) relationships that degrade emulation speed and increase hardware utilization, we propose a device and variability modeling approach based on device measurements. We deploy look-up tables (LUTs) for device modeling and use the multivariate kernel density estimation (KDE) method to augment existing data, extending data variety and avoiding repetitive data usage. The proposed emulator achieves cycle-accurate, real-time emulations and provides information such as latency and energy consumption for matrix mapping and vector-matrix multiplications (VMMs). Experimental results show a significant reduction in emulation time compared to conventional behavioral simulations. Additionally, an RRAM-based discrete Fourier transform (DFT) accelerator is analyzed as a case study featuring a range of in-depth system assessments.}, language = {en} }