cs.AR

arXiv:2304.12467v3 Announce Type: replace Abstract: Neural Radiance Field (NeRF) based 3D reconstruction is highly desirable for immersive Augmented and Virtual Reality (AR/VR) applications, but achieving instant (i.e., < 5 seconds) on-device NeRF training remains a challenge. In this work, we first identify the inefficiency bottleneck: the need to interpolate NeRF embeddings up to 200,000 times from a 3D embedding grid during each training iteration. To alleviate this, we propose Instant-3D, an algorithm-hardware co-design acceleration framework that achieves instant on-device NeRF training. Our algorithm decomposes the embedding grid representation in terms of color and density, enabling computational redundancy to be squeezed out by adopting different (1) grid sizes and (2) update frequencies for the color and density branches. Our hardware accelerator further reduces the dominant memory accesses for embedding grid interpolation by (1) mapping multiple nearby points' memory read requests into one during the feed-forward process, (2) merging embedding grid updates from the same sliding time window during back-propagation, and (3) fusing different computation cores to support the different grid sizes needed by the color and density branches of Instant-3D algorithm. Extensive experiments validate the effectiveness of Instant-3D, achieving a large training time reduction of 41x - 248x while maintaining the same reconstruction quality. Excitingly, Instant-3D has enabled instant 3D reconstruction for AR/VR, requiring a reconstruction time of only 1.6 seconds per scene and meeting the AR/VR power consumption constraint of 1.9 W.

arXiv:2503.22567v1 Announce Type: new Abstract: Efficient on-device neural network (NN) inference has various advantages over cloud-based processing, including predictable latency, enhanced privacy, greater reliability, and reduced operating costs for vendors. This has sparked the recent rapid development of microcontroller-scale NN accelerators, often referred to as neural processing units ($\mu$NPUs), designed specifically for ultra-low-power applications. In this paper we present the first comparative evaluation of a number of commercially-available $\mu$NPUs, as well as the first independent benchmarks for several of these platforms. We develop and open-source a model compilation framework to enable consistent benchmarking of quantized models across diverse $\mu$NPU hardware. Our benchmark targets end-to-end performance and includes model inference latency, power consumption, and memory overhead, alongside other factors. The resulting analysis uncovers both expected performance trends as well as surprising disparities between hardware specifications and actual performance, including $\mu$NPUs exhibiting unexpected scaling behaviors with increasing model complexity. Our framework provides a foundation for further evaluation of $\mu$NPU platforms alongside valuable insights for both hardware designers and software developers in this rapidly evolving space.

arXiv:2206.08141v3 Announce Type: replace Abstract: We present a first-of-its-kind ultra-compact intelligent camera system, dubbed i-FlatCam, including a lensless camera with a computational (Comp.) chip. It highlights (1) a predict-then-focus eye tracking pipeline for boosted efficiency without compromising the accuracy, (2) a unified compression scheme for single-chip processing and improved frame rate per second (FPS), and (3) dedicated intra-channel reuse design for depth-wise convolutional layers (DW-CONV) to increase utilization. i-FlatCam demonstrates the first eye tracking pipeline with a lensless camera and achieves 3.16 degrees of accuracy, 253 FPS, 91.49 $\mu$J/Frame, and 6.7mm x 8.9mm x 1.2mm camera form factor, paving the way for next-generation Augmented Reality (AR) and Virtual Reality (VR) devices.

arXiv:2503.22044v1 Announce Type: new Abstract: Compute-in-memory (CIM) based neural network accelerators offer a promising solution to the Von Neumann bottleneck by computing directly within memory arrays. However, SRAM CIM faces limitations in executing larger models due to its cell size and on-chip memory constraints. This work proposes CIMPool, a CIM-aware compression and acceleration framework that counters this limitation through a weight sharing-based compression technique, aptly named `Weight Pool,' enabling significantly larger neural networks to be accommodated within on-chip memory constraints. This method minimizes the accuracy trade-off typically associated with parameter compression, allowing CIMPool to achieve a significantly larger compression ratio compared to the traditional quantization method with iso-accuracy. Furthermore, CIMPool co-optimizes the compression algorithm, hardware, and dataflow to efficiently implement the hardware permutation required by weight pool compression, with negligible area and throughput overhead. Empirical results demonstrate that CIMPool can achieve 8-bit level accuracy with an effective 0.5-bit precision, reduce chip area by 62.3% for ResNet-18, and enable the execution of an order of magnitude larger models for a given area budget in SRAM CIMs. When DRAM is used to store weights, CIMPool can reduce the total energy by 3.24x compared to iso-accuracy traditional CIMs.

arXiv:2503.22072v1 Announce Type: new Abstract: Computing-in-memory (CIM) is renowned in deep learning due to its high energy efficiency resulting from highly parallel computing with minimal data movement. However, current SRAM-based CIM designs suffer from long latency for loading weight or feature maps from DRAM for large AI models. Moreover, previous SRAM-based CIM architectures lack end-to-end model inference. To address these issues, this paper proposes CIMR-V, an end-to-end CIM accelerator with RISC-V that incorporates CIM layer fusion, convolution/max pooling pipeline, and weight fusion, resulting in an 85.14\% reduction in latency for the keyword spotting model. Furthermore, the proposed CIM-type instructions facilitate end-to-end AI model inference and full stack flow, effectively synergizing the high energy efficiency of CIM and the high programmability of RISC-V. Implemented using TSMC 28nm technology, the proposed design achieves an energy efficiency of 3707.84 TOPS/W and 26.21 TOPS at 50 MHz.

arXiv:2503.22017v1 Announce Type: new Abstract: The growing prevalence of data-intensive workloads, such as artificial intelligence (AI), machine learning (ML), high-performance computing (HPC), in-memory databases, and real-time analytics, has exposed limitations in conventional memory technologies like DRAM. While DRAM offers low latency and high throughput, it is constrained by high costs, scalability challenges, and volatility, making it less viable for capacity-bound and persistent applications in modern datacenters. Recently, Compute Express Link (CXL) has emerged as a promising alternative, enabling high-speed, cacheline-granular communication between CPUs and external devices. By leveraging CXL technology, NAND flash can now be used as memory expansion, offering three-fold benefits: byte-addressability, scalable capacity, and persistence at a low cost. Samsung's CXL Memory Module Hybrid (CMM-H) is the first product to deliver these benefits through a hardware-only solution, i.e., it does not incur any OS and IO overheads like conventional block devices. In particular, CMM-H integrates a DRAM cache with NAND flash in a single device to deliver near-DRAM latency. This paper presents the first publicly available study for comprehensive characterizations of an FPGA-based CMM-H prototype. Through this study, we address users' concerns about whether a wide variety of applications can successfully run on a memory device backed by NAND flash medium. Additionally, based on these characterizations, we provide key insights into how to best take advantage of the CMM-H device.

arXiv:2412.14029v2 Announce Type: replace-cross Abstract: In analog neuromorphic chips, designers can embed computing primitives in the intrinsic physical properties of devices and circuits, heavily reducing device count and energy consumption, and enabling high parallelism, because all devices are computing simultaneously. Neural network parameters can be stored in local analog non-volatile memories (NVMs), saving the energy required to move data between memory and logic. However, the main drawback of analog sub-threshold electronic circuits is their dramatic temperature sensitivity. In this paper, we demonstrate that a temperature compensation mechanism can be devised to solve this problem. We have designed and fabricated a chip implementing a two-layer analog neural network trained to classify low-resolution images of handwritten digits with a low-cost single-poly complementary metal-oxide-semiconductor (CMOS) process, using unconventional analog NVMs for weight storage. We demonstrate a temperature-resilient analog neuromorphic chip for image recognition operating between 10$^{\circ}$C and 60$^{\circ}$C without loss of classification accuracy, within 2\% of the corresponding software-based neural network in the whole temperature range.

arXiv:2212.01120v2 Announce Type: replace Abstract: Neural Radiance Field (NeRF) based rendering has attracted growing attention thanks to its state-of-the-art (SOTA) rendering quality and wide applications in Augmented and Virtual Reality (AR/VR). However, immersive real-time (> 30 FPS) NeRF based rendering enabled interactions are still limited due to the low achievable throughput on AR/VR devices. To this end, we first profile SOTA efficient NeRF algorithms on commercial devices and identify two primary causes of the aforementioned inefficiency: (1) the uniform point sampling and (2) the dense accesses and computations of the required embeddings in NeRF. Furthermore, we propose RT-NeRF, which to the best of our knowledge is the first algorithm-hardware co-design acceleration of NeRF. Specifically, on the algorithm level, RT-NeRF integrates an efficient rendering pipeline for largely alleviating the inefficiency due to the commonly adopted uniform point sampling method in NeRF by directly computing the geometry of pre-existing points. Additionally, RT-NeRF leverages a coarse-grained view-dependent computing ordering scheme for eliminating the (unnecessary) processing of invisible points. On the hardware level, our proposed RT-NeRF accelerator (1) adopts a hybrid encoding scheme to adaptively switch between a bitmap- or coordinate-based sparsity encoding format for NeRF's sparse embeddings, aiming to maximize the storage savings and thus reduce the required DRAM accesses while supporting efficient NeRF decoding; and (2) integrates both a dual-purpose bi-direction adder & search tree and a high-density sparse search unit to coordinate the two aforementioned encoding formats. Extensive experiments on eight datasets consistently validate the effectiveness of RT-NeRF, achieving a large throughput improvement (e.g., 9.7x - 3,201x) while maintaining the rendering quality as compared with SOTA efficient NeRF solutions.

arXiv:2503.07242v2 Announce Type: replace Abstract: Convolutional Neural Networks (CNNs) serve various applications with diverse performance and resource requirements. Model-aware CNN accelerators best address these diverse requirements. These accelerators usually combine multiple dedicated Compute Engines (CEs). The flexibility of Field-Programmable Gate Arrays (FPGAs) enables the design of such multiple Compute-Engine (multiple-CE) accelerators. However, existing multiple-CE accelerators differ in how they arrange their CEs and distribute the FPGA resources and CNN operators among the CEs. The design space of multiple-CE accelerators comprises numerous such arrangements, which makes a systematic identification of the best ones an open challenge. This paper proposes a multiple-CE accelerator analytical Cost Model (MCCM) and an evaluation methodology built around MCCM. The model and methodology streamline the expression of any multiple-CE accelerator and provide a fast evaluation of its performance and efficiency. MCCM is in the order of 100000x faster than traditional synthesis-based evaluation and has an average accuracy of > 90%. The paper presents three use cases of MCCM. The first describes an end-to-end evaluation of state-of-the-art multiple-CE accelerators considering various metrics, CNN models, and resource budgets. The second describes fine-grained evaluation that helps identify performance bottlenecks of multiple-CE accelerators. The third demonstrates that MCCM fast evaluation enables exploring the vast design space of multiple-CE accelerators. These use cases show that no unique CE arrangement achieves the best results given different metrics, CNN models, and resource budgets. They also show that fast evaluation enables design space exploration, resulting in accelerator designs that outperform state-of-the-art ones. MCCM is available at https://github.com/fqararyah/MCCM.

arXiv:2503.09676v1 Announce Type: cross Abstract: Research into the development of special-purpose computing architectures designed to solve quadratic unconstrained binary optimization (QUBO) problems has flourished in recent years. It has been demonstrated in the literature that such special-purpose solvers can outperform traditional CMOS architectures by orders of magnitude with respect to timing metrics on synthetic problems. However, they face challenges with constrained problems such as the quadratic assignment problem (QAP), where mapping to binary formulations such as QUBO introduces overhead and limits parallelism. In-memory computing (IMC) devices, such as memristor-based analog Ising machines, offer significant speedups and efficiency gains over traditional CPU-based solvers, particularly for solving combinatorial optimization problems. In this work, we present a novel local search heuristic designed for IMC hardware to tackle the QAP. Our approach enables massive parallelism that allows for computing of full neighbourhoods simultaneously to make update decisions. We ensure binary solutions remain feasible by selecting local moves that lead to neighbouring feasible solutions, leveraging feasible-space search heuristics and the underlying structure of a given problem. Our approach is compatible with both digital computers and analog hardware. We demonstrate its effectiveness in CPU implementations by comparing it with state-of-the-art heuristics for solving the QAP.

arXiv:2502.07842v2 Announce Type: replace Abstract: Compute-in-memory (CIM) is an efficient method for implementing deep neural networks (DNNs) but suffers from substantial overhead from analog-to-digital converters (ADCs), especially as ADC precision increases. Low-precision ADCs can reduce this overhead but introduce partial-sum quantization errors degrading accuracy. Additionally, low-bit weight constraints, imposed by cell limitations and the need for multiple cells for higher-bit weights, present further challenges. While fine-grained partial-sum quantization has been studied to lower ADC resolution effectively, weight granularity, which limits overall partial-sum quantized accuracy, remains underexplored. This work addresses these challenges by aligning weight and partial-sum quantization granularities at the column-wise level. Our method improves accuracy while maintaining dequantization overhead, simplifies training by removing two-stage processes, and ensures robustness to memory cell variations via independent column-wise scale factors. We also propose an open-source CIM-oriented convolution framework to handle fine-grained weights and partial-sums efficiently, incorporating a novel tiling method and group convolution. Experimental results on ResNet-20 (CIFAR-10, CIFAR-100) and ResNet-18 (ImageNet) show accuracy improvements of 0.99%, 2.69%, and 1.01%, respectively, compared to the best-performing related works. Additionally, variation analysis reveals the robustness of our method against memory cell variations. These findings highlight the effectiveness of our quantization scheme in enhancing accuracy and robustness while maintaining hardware efficiency in CIM-based DNN implementations. Our code is available at https://github.com/jiyoonkm/ColumnQuant.

arXiv:2503.10296v1 Announce Type: new Abstract: This paper discusses the integration challenges and strategies for designing mobile robots, by focusing on the task-driven, optimal selection of hardware and software to balance safety, efficiency, and minimal usage of resources such as costs, energy, computational requirements, and weight. We emphasize the interplay between perception and motion planning in decision-making by introducing the concept of occupancy queries to quantify the perception requirements for sampling-based motion planners. Sensor and algorithm performance are evaluated using False Negative Rates (FPR) and False Positive Rates (FPR) across various factors such as geometric relationships, object properties, sensor resolution, and environmental conditions. By integrating perception requirements with perception performance, an Integer Linear Programming (ILP) approach is proposed for efficient sensor and algorithm selection and placement. This forms the basis for a co-design optimization that includes the robot body, motion planner, perception pipeline, and computing unit. We refer to this framework for solving the co-design problem of mobile robots as CODEI, short for Co-design of Embodied Intelligence. A case study on developing an Autonomous Vehicle (AV) for urban scenarios provides actionable information for designers, and shows that complex tasks escalate resource demands, with task performance affecting choices of the autonomy stack. The study demonstrates that resource prioritization influences sensor choice: cameras are preferred for cost-effective and lightweight designs, while lidar sensors are chosen for better energy and computational efficiency.

arXiv:2503.10207v1 Announce Type: new Abstract: Kyber, an IND-CCA2-secure lattice-based post-quantum key-encapsulation mechanism, is the winner of the first post-quantum cryptography standardization process of the US National Institute of Standards and Technology. In this work, we provide an efficient implementation of Kyber on ESP32, a very popular microcontroller for Internet of Things applications. We hand-partition the Kyber algorithm to enable utilization of the ESP32 dual-core architecture, which allows us to speed up its execution by 1.21x (keygen), 1.22x (encaps) and 1.20x (decaps). We also explore the possibility of gaining further improvement by utilizing the ESP32 SHA and AES coprocessor and achieve a culminated speed-up of 1.72x (keygen), 1.84x (encaps) and 1.69x (decaps).

arXiv:2503.09650v1 Announce Type: new Abstract: With the advancement of Large Language Models (LLMs), the importance of accelerators that efficiently process LLM computations has been increasing. This paper discusses the necessity of LLM accelerators and provides a comprehensive analysis of the hardware and software characteristics of the main commercial LLM accelerators. Based on this analysis, we propose considerations for the development of next-generation LLM accelerators and suggest future research directions.

arXiv:2404.18407v2 Announce Type: replace Abstract: Physical design watermarking on contemporary integrated circuit (IC) layout encodes signatures without considering the dense connections and design constraints, which could lead to performance degradation on the watermarked products. This paper presents ICMarks, a quality-preserving and robust watermarking framework for modern IC physical design. ICMarks embeds unique watermark signatures during the physical design's placement stage, thereby authenticating the IC layout ownership. ICMarks's novelty lies in (i) strategically identifying a region of cells to watermark with minimal impact on the layout performance and (ii) a two-level watermarking framework for augmented robustness toward potential removal and forging attacks. Extensive evaluations on benchmarks of different design objectives and sizes validate that ICMarks incurs no wirelength and timing metrics degradation, while successfully proving ownership. Furthermore, we demonstrate ICMarks is robust against two major watermarking attack categories, namely, watermark removal and forging attacks; even if the adversaries have prior knowledge of the watermarking schemes, the signatures cannot be removed without significantly undermining the layout quality.

arXiv:2503.09975v1 Announce Type: new Abstract: Low-precision data types are essential in modern neural networks during both training and inference as they enhance throughput and computational capacity by better exploiting available hardware resources. Despite the incorporation of FP8 in commercially available neural network accelerators, a comprehensive exposition of its underlying mechanisms, along with rigorous performance and accuracy evaluations, is still lacking. In this work, we contribute in three significant ways. First, we analyze the implementation details and quantization options associated with FP8 for inference on the Intel Gaudi AI accelerator. Second, we empirically quantify the throughput improvements afforded by the use of FP8 at both the operator level and in end-to-end scenarios. Third, we assess the accuracy impact of various FP8 quantization methods. Our experimental results indicate that the Intel Gaudi 2 accelerator consistently achieves high computational unit utilization, frequently exceeding 90\% MFU, while incurring an accuracy degradation of less than 1\%.

arXiv:2502.20415v2 Announce Type: replace Abstract: Neuromorphic computing is a relatively new discipline of computer science, where the principles of biological brain's computation and memory are used to create a new way of processing information, based on networks of spiking neurons. Those networks can be implemented as both analog and digital implementations, where for the latter, the Field Programmable Gate Arrays (FPGAs) are a frequent choice, due to their inherent flexibility, allowing the researchers to easily design hardware neuromorphic architecture (NMAs). Moreover, digital NMAs show good promise in simulating various spiking neural networks because of their inherent accuracy and resilience to noise, as opposed to analog implementations. This paper presents an overview of digital NMAs implemented on FPGAs, with a goal of providing useful references to various architectural design choices to the researchers interested in digital neuromorphic systems. We present a taxonomy of NMAs that highlights groups of distinct architectural features, their advantages and disadvantages and identify trends and predictions for the future of those architectures.

arXiv:2307.05815v2 Announce Type: replace Abstract: The globalization of semiconductor supply chains has exposed Network-on-Chip (NoC) interconnects in System-on-Chip (SoC) architectures to critical security risks, including reverse engineering and IP theft. To address these threats, this work builds on two methodologies: ObNoCs [11], which obfuscates NoC topologies using programmable multiplexers, and POTENT [10], which enhances post-synthesis security against SAT-based attacks. These techniques ensure robust protection of NoC interconnects with minimal performance overhead. As the industry shifts to chiplet-based heterogeneous architectures, this research extends ObNoCs and POTENT to secure intra- and inter-chiplet interconnects. New challenges, such as safeguarding inter-chiplet communication and interposer design, are addressed through enhanced obfuscation, authentication, and encryption mechanisms. Experimental results demonstrate the practicality of these approaches for high-security applications, ensuring trust and reliability in monolithic and modular systems.

arXiv:2503.05116v1 Announce Type: new Abstract: Graph processing requires irregular, fine-grained random access patterns incompatible with contemporary off-chip memory architecture, leading to inefficient data access. This inefficiency makes graph processing an extremely memory-bound application. Because of this, existing graph processing accelerators typically employ a graph tiling-based or processing-in-memory (PIM) approach to relieve the memory bottleneck. In the tiling-based approach, a graph is split into chunks that fit within the on-chip cache to maximize data reuse. In the PIM approach, arithmetic units are placed within memory to perform operations such as reduction or atomic addition. However, both approaches have several limitations, especially when implemented on current memory standards (i.e., DDR). Because the access granularity provided by DDR is much larger than that of the graph vertex property data, much of the bandwidth and cache capacity are wasted. PIM is meant to alleviate such issues, but it is difficult to use in conjunction with the tiling-based approach, resulting in a significant disadvantage. Furthermore, placing arithmetic units inside a memory chip is expensive, thereby supporting multiple types of operation is thought to be impractical. To address the above limitations, we present Piccolo, an end-to-end efficient graph processing accelerator with fine-grained in-memory random scatter-gather. Instead of placing expensive arithmetic units in off-chip memory, Piccolo focuses on reducing the off-chip traffic with non-arithmetic function-in-memory of random scatter-gather. To fully benefit from in-memory scatter-gather, Piccolo redesigns the cache and MHA of the accelerator such that it can enjoy both the advantage of tiling and in-memory operations. Piccolo achieves a maximum speedup of 3.28$\times$ and a geometric mean speedup of 1.62$\times$ across various and extensive benchmarks.

arXiv:2503.05197v1 Announce Type: new Abstract: Point clouds are increasingly important in intelligent applications, but frequent off-chip memory traffic in accelerators causes pipeline stalls and leads to high energy consumption. While conventional line buffer techniques can eliminate off-chip traffic, they cannot be directly applied to point clouds due to their inherent computation patterns. To address this, we introduce two techniques: compulsory splitting and deterministic termination, enabling fully-streaming processing. We further propose StreamGrid, a framework that integrates these techniques and automatically optimizes on-chip buffer sizes. Our evaluation shows StreamGrid reduces on-chip memory by 61.3\% and energy consumption by 40.5\% with marginal accuracy loss compared to the baselines without our techniques. Additionally, we achieve 10.0$\times$ speedup and 3.9$\times$ energy efficiency over state-of-the-art accelerators.