Searcharxiv⌕ Search

arXiv subjects

Luhong Liang

Publications and source records attributed to Luhong Liang.

8 recordsLinked to original sources
arXiv2026

31.1 A 14.08-to-135.69Token/s ReRAM-on-Logic Stacked Outlier-Free Large-Language-Model Accelerator with Block-Clustered Weight-Compression and Adaptive Parallel-Speculative-Decoding

This work presents a 55nm speculative decoding-based LLM accelerator with bumping-based face-to-face ReRAM-on-logic stacking technology. It features a local rotation unit for outlier-free low-bit quantization, a stacking-aware PNM architecture co-designed with blockwise vector quantization to reduce weight EMA overheads, and an adaptive parallel speculative decoding scheme with an out-of-order scheduler for high resource and bandwidth utilization. Our chip achieves 14.08-to-135.69token/s and 4.46-to-7.17x speedup over vanilla speculative decoding.

cs.AR↗
arXiv2026

Expert Streaming: Accelerating Low-Batch MoE Inference via Multi-chiplet Architecture and Dynamic Expert Trajectory Scheduling

Mixture-of-Experts is a promising approach for edge AI with low-batch inference. Yet, on-device deployments often face limited on-chip memory and severe workload imbalance; the prevalent use of offloading further incurs off-chip memory access bottlenecks. Moreover, MoE sparsity and dynamic gating shift distributed strategies toward much finer granularity and introduce runtime scheduling considerations. Recently, high die-to-die bandwidth chiplet interconnects have created new opportunities for multi-chiplet systems to address workload imbalance and offloading bottlenecks with fine-grained scheduling. In this paper, we propose Fully Sharded Expert Data Parallelism, a parallelization paradigm specifically architected for low-batch MoE inference on multi-chiplet accelerators. FSE-DP attains adaptive computation-communication overlap and balanced load by orchestrating fine-grained, complementary expert streams along dynamic trajectories across high-bandwidth D2D links. The attendant dataflow complexity is tamed by a minimal, hardware-amenable set of virtualization rules and a lightweight scheduling algorithm. Our approach achieves 1.22 to 2.00 times speedup over state-of-the-art baselines and saves up to 78.8 percent on-chip memory.

cs.AR↗
arXiv2026

Towards Secure and Efficient DNN Accelerators via Hardware-Software Co-Design

The rapid deployment of deep neural network (DNN) accelerators in safety-critical domains such as autonomous vehicles, healthcare systems, and financial infrastructure necessitates robust mechanisms to safeguard data confidentiality and computational integrity. Existing security solutions for DNN accelerators, however, suffer from excessive hardware resource demands and frequent off-chip memory access overheads, which degrade performance and scalability. To address these challenges, this paper presents a secure and efficient memory protection framework for DNN accelerators with minimal overhead. First, we propose a bandwidth-aware cryptographic scheme that adapts encryption granularity based on memory traffic patterns, striking a balance between security and resource efficiency. Second, we observe that both the overlapping regions in the intra-layer tiling's sliding window pattern and those resulting from inter-layer tiling strategy discrepancies introduce substantial redundant memory accesses and repeated computational overhead in cryptography. Third, we introduce a multi-level authentication mechanism that effectively eliminates unnecessary off-chip memory accesses, enhancing performance and energy efficiency. Experimental results show that this work decreases performance overhead by over 12% and achieves 87% energy efficiency improvement for both server and edge neural processing units (NPUs), while ensuring robust scalability.

cs.CR↗
arXiv2025

SeDA: Secure and Efficient DNN Accelerators with Hardware/Software Synergy

Ensuring the confidentiality and integrity of DNN accelerators is paramount across various scenarios spanning autonomous driving, healthcare, and finance. However, current security approaches typically require extensive hardware resources, and incur significant off-chip memory access overheads. This paper introduces SeDA, which utilizes 1) a bandwidth-aware encryption mechanism to improve hardware resource efficiency, 2) optimal block granularity through intra-layer and inter-layer tiling patterns, and 3) a multi-level integrity verification mechanism that minimizes, or even eliminates, memory access overheads. Experimental results show that SeDA decreases performance overhead by over 12% for both server and edge neural processing units (NPUs), while ensuring robust scalability.

cs.AR↗
arXiv2025

APSQ: Additive Partial Sum Quantization with Algorithm-Hardware Co-Design

DNN accelerators, significantly advanced by model compression and specialized dataflow techniques, have marked considerable progress. However, the frequent access of high-precision partial sums (PSUMs) leads to excessive memory demands in architectures utilizing input/weight stationary dataflows. Traditional compression strategies have typically overlooked PSUM quantization, which may account for 69% of power consumption. This study introduces a novel Additive Partial Sum Quantization (APSQ) method, seamlessly integrating PSUM accumulation into the quantization framework. A grouping strategy that combines APSQ with PSUM quantization enhanced by a reconfigurable architecture is further proposed. The APSQ performs nearly lossless on NLP and CV tasks across BERT, Segformer, and EfficientViT models while compressing PSUMs to INT8. This leads to a notable reduction in energy costs by 28-87%. Extended experiments on LLaMA2-7B demonstrate the potential of APSQ for large language models. Code is available at https://github.com/Yonghao-Tan/APSQ.

cs.AR↗
arXiv2024

A 10.60 $μ$W 150 GOPS Mixed-Bit-Width Sparse CNN Accelerator for Life-Threatening Ventricular Arrhythmia Detection

This paper proposes an ultra-low power, mixed-bit-width sparse convolutional neural network (CNN) accelerator to accelerate ventricular arrhythmia (VA) detection. The chip achieves 50% sparsity in a quantized 1D CNN using a sparse processing element (SPE) architecture. Measurement on the prototype chip TSMC 40nm CMOS low-power (LP) process for the VA classification task demonstrates that it consumes 10.60 $μ$W of power while achieving a performance of 150 GOPS and a diagnostic accuracy of 99.95%. The computation power density is only 0.57 $μ$W/mm$^2$, which is 14.23X smaller than state-of-the-art works, making it highly suitable for implantable and wearable medical devices.

cs.AR↗
arXiv2024

Genetic Quantization-Aware Approximation for Non-Linear Operations in Transformers

Non-linear functions are prevalent in Transformers and their lightweight variants, incurring substantial and frequently underestimated hardware costs. Previous state-of-the-art works optimize these operations by piece-wise linear approximation and store the parameters in look-up tables (LUT), but most of them require unfriendly high-precision arithmetics such as FP/INT 32 and lack consideration of integer-only INT quantization. This paper proposed a genetic LUT-Approximation algorithm namely GQA-LUT that can automatically determine the parameters with quantization awareness. The results demonstrate that GQA-LUT achieves negligible degradation on the challenging semantic segmentation task for both vanilla and linear Transformer models. Besides, proposed GQA-LUT enables the employment of INT8-based LUT-Approximation that achieves an area savings of 81.3~81.7% and a power reduction of 79.3~80.2% compared to the high-precision FP/INT 32 alternatives. Code is available at https:// github.com/PingchengDong/GQA-LUT.

cs.LG↗