arXiv subjects

Luciano Lavagno

Publications and source records attributed to Luciano Lavagno.

8 recordsLinked to original sources

arXiv2025

eIQ Neutron: Redefining Edge-AI Inference with Integrated NPU and Compiler Innovations

Neural Processing Units (NPUs) are key to enabling efficient AI inference in resource-constrained edge environments. While peak tera operations per second (TOPS) is often used to gauge performance, it poorly reflects real-world performance and typically rather correlates with higher silicon cost. To address this, architects must focus on maximizing compute utilization, without sacrificing flexibility. This paper presents the eIQ Neutron efficient-NPU, integrated into a commercial flagship MPU, alongside co-designed compiler algorithms. The architecture employs a flexible, data-driven design, while the compiler uses a constrained programming approach to optimize compute and data movement based on workload characteristics. Compared to the leading embedded NPU and compiler stack, our solution achieves an average speedup of 1.8x (4x peak) at equal TOPS and memory resources across standard AI-benchmarks. Even against NPUs with double the compute and memory resources, Neutron delivers up to 3.3x higher performance.

cs.AR↗

arXiv2025

Just TestIt! An SBST Approach To Automate System-Integration Testing

This paper introduces TestIt, an open-source Python package designed to automate full-system integration testing using a Software-Based Self-Test (SBST) approach. By dynamically generating test vectors and golden references, TestIt significantly reduces development time and complexity while supporting both simulation and FPGA environments. Its flexible design positions TestIt as a key enabler for the widespread adoption of CI/CD methodologies in open-source RTL development. A case study on the X-HEEP RISC-V microcontroller (MCU), which integrates a custom accelerator, showcases TestIt's ability to detect hardware and software faults that traditional formal methods may overlook. Furthermore, the case study highlights how TestIt can be leveraged to characterize system performance with minimal effort. By automating testing on the PYNQ-Z2 FPGA development board, we achieved a 11x speed-up with respect to RTL simulations.

cs.AR↗

arXiv2025

A Survey on Deep Learning Hardware Accelerators for Heterogeneous HPC Platforms

Recent trends in deep learning (DL) have made hardware accelerators essential for various high-performance computing (HPC) applications, including image classification, computer vision, and speech recognition. This survey summarizes and classifies the most recent developments in DL accelerators, focusing on their role in meeting the performance demands of HPC applications. We explore cutting-edge approaches to DL acceleration, covering not only GPU- and TPU-based platforms but also specialized hardware such as FPGA- and ASIC-based accelerators, Neural Processing Units, open hardware RISC-V-based accelerators, and co-processors. This survey also describes accelerators leveraging emerging memory technologies and computing paradigms, including 3D-stacked Processor-In-Memory, non-volatile memories like Resistive RAM and Phase Change Memories used for in-memory computing, as well as Neuromorphic Processing Units, and Multi-Chip Module-based accelerators. Furthermore, we provide insights into emerging quantum-based accelerators and photonics. Finally, this survey categorizes the most influential architectures and technologies from recent years, offering readers a comprehensive perspective on the rapidly evolving field of deep learning acceleration.

cs.AR↗

arXiv2024

SILVIA: Automated Superword-Level Parallelism Exploitation via HLS-Specific LLVM Passes for Compute-Intensive FPGA Accelerators

High-level synthesis (HLS) aims at democratizing custom hardware acceleration with highly abstracted software-like descriptions. However, efficient accelerators still require substantial low-level hardware optimizations, defeating the HLS intent. In the context of field-programmable gate arrays, digital signal processors (DSPs) are a crucial resource that typically requires a significant optimization effort for its efficient utilization, especially when used for sub-word vectorization. This work proposes SILVIA, an open-source LLVM transformation pass that automatically identifies superword-level parallelism within an HLS design and exploits it by packing multiple operations, such as additions, multiplications, and multiply-and-adds, into a single DSP. SILVIA is integrated in the flow of the commercial AMD Vitis HLS tool and proves its effectiveness by packing multiple operations on the DSPs without any manual source-code modifications on several diverse state-of-the-art HLS designs such as convolutional neural networks and basic linear algebra subprograms accelerators, reducing the DSP utilization for additions by 70 % and for multiplications and multiply-and-adds by 50 % on average.

cs.AR↗

arXiv2024

To Spike or Not To Spike: A Digital Hardware Perspective on Deep Learning Acceleration

As deep learning models scale, they become increasingly competitive from domains spanning from computer vision to natural language processing; however, this happens at the expense of efficiency since they require increasingly more memory and computing power. The power efficiency of the biological brain outperforms any large-scale deep learning ( DL ) model; thus, neuromorphic computing tries to mimic the brain operations, such as spike-based information processing, to improve the efficiency of DL models. Despite the benefits of the brain, such as efficient information transmission, dense neuronal interconnects, and the co-location of computation and memory, the available biological substrate has severely constrained the evolution of biological brains. Electronic hardware does not have the same constraints; therefore, while modeling spiking neural networks ( SNNs) might uncover one piece of the puzzle, the design of efficient hardware backends for SNN s needs further investigation, potentially taking inspiration from the available work done on the artificial neural networks ( ANNs) side. As such, when is it wise to look at the brain while designing new hardware, and when should it be ignored? To answer this question, we quantitatively compare the digital hardware acceleration techniques and platforms of ANNs and SNN s. As a result, we provide the following insights: (i) ANNs currently process static data more efficiently, (ii) applications targeting data produced by neuromorphic sensors, such as event-based cameras and silicon cochleas, need more investigation since the behavior of these sensors might naturally fit the SNN paradigm, and (iii) hybrid approaches combining SNN s and ANNs might lead to the best solutions and should be investigated further at the hardware level, accounting for both efficiency and loss optimization.

cs.NE↗

arXiv2023

Design and Optimization of Residual Neural Network Accelerators for Low-Power FPGAs Using High-Level Synthesis

Residual neural networks are widely used in computer vision tasks. They enable the construction of deeper and more accurate models by mitigating the vanishing gradient problem. Their main innovation is the residual block which allows the output of one layer to bypass one or more intermediate layers and be added to the output of a later layer. Their complex structure and the buffering required by the residual block make them difficult to implement on resource-constrained platforms. We present a novel design flow for implementing deep learning models for field programmable gate arrays optimized for ResNets, using a strategy to reduce their buffering overhead to obtain a resource-efficient implementation of the residual layer. Our high-level synthesis (HLS)-based flow encompasses a thorough set of design principles and optimization strategies, exploiting in novel ways standard techniques such as temporal reuse and loop merging to efficiently map ResNet models, and potentially other skip connection-based NN architectures, into FPGA. The models are quantized to 8-bit integers for both weights and activations, 16-bit for biases, and 32-bit for accumulations. The experimental results are obtained on the CIFAR-10 dataset using ResNet8 and ResNet20 implemented with Xilinx FPGAs using HLS on the Ultra96-V2 and Kria KV260 boards. Compared to the state-of-the-art on the Kria KV260 board, our ResNet20 implementation achieves 2.88X speedup with 0.5% higher accuracy of 91.3%, while ResNet8 accuracy improves by 2.8% to 88.7%. The throughputs of ResNet8 and ResNet20 are 12971 FPS and 3254 FPS on the Ultra96 board, and 30153 FPS and 7601 FPS on the Kria KV26, respectively. They Pareto-dominate state-of-the-art solutions concerning accuracy, throughput, and energy.

cs.AR↗

arXiv2023

A DSP shared is a DSP earned: HLS Task-Level Multi-Pumping for High-Performance Low-Resource Designs

High-level synthesis (HLS) enhances digital hardware design productivity through a high abstraction level. Even if the HLS abstraction prevents fine-grained manual register-transfer level (RTL) optimizations, it also enables automatable optimizations that would be unfeasible or hard to automate at RTL. Specifically, we propose a task-level multi-pumping methodology to reduce resource utilization, particularly digital signal processors (DSPs), while preserving the throughput of HLS kernels modeled as dataflow graphs (DFGs) targeting field-programmable gate arrays. The methodology exploits the HLS resource sharing to automatically insert the logic for reusing the same functional unit for different operations. In addition, it relies on multi-clock DFG s to run the multi-pumped tasks at higher frequencies. The methodology scales the pipeline initiation interval (II) and the clock frequency constraints of resource-intensive tasks by a multi-pumping factor (M). The looser II allows sharing the same resource among M different operations, while the tighter clock frequency preserves the throughput. We verified that our methodology opens a new Pareto front in the throughput and resource space by applying it to open-source HLS designs using state-of-the-art commercial HLS and implementation tools by Xilinx. The multi-pumped designs require up to 40% fewer DSP resources at the same throughput as the original designs optimized for performance (i.e., running at the maximum clock frequency) and achieve up to 50% better throughput using the same DSP s as the original designs optimized for resources with a single clock.

cs.AR↗

arXiv2020

Synetgy: Algorithm-hardware Co-design for ConvNet Accelerators on Embedded FPGAs

Using FPGAs to accelerate ConvNets has attracted significant attention in recent years. However, FPGA accelerator design has not leveraged the latest progress of ConvNets. As a result, the key application characteristics such as frames-per-second (FPS) are ignored in favor of simply counting GOPs, and results on accuracy, which is critical to application success, are often not even reported. In this work, we adopt an algorithm-hardware co-design approach to develop a ConvNet accelerator called Synetgy and a novel ConvNet model called DiracDeltaNet$^{\dagger}$. Both the accelerator and ConvNet are tailored to FPGA requirements. DiracDeltaNet, as the name suggests, is a ConvNet with only $1\times 1$ convolutions while spatial convolutions are replaced by more efficient shift operations. DiracDeltaNet achieves competitive accuracy on ImageNet (88.7\% top-5), but with 42$\times$ fewer parameters and 48$\times$ fewer OPs than VGG16. We further quantize DiracDeltaNet's weights to 4-bit and activations to 4-bits, with less than 1\% accuracy loss. These quantizations exploit well the nature of FPGA hardware. In short, DiracDeltaNet's small model size, low computational OP count, low precision and simplified operators allow us to co-design a highly customized computing unit for an FPGA. We implement the computing units for DiracDeltaNet on an Ultra96 SoC system through high-level synthesis. Our accelerator's final top-5 accuracy of 88.1\% on ImageNet, is higher than all the previously reported embedded FPGA accelerators. In addition, the accelerator reaches an inference speed of 66.3 FPS on the ImageNet classification task, surpassing prior works with similar accuracy by at least 11.6$\times$.

cs.CV↗