Large Language Models (LLMs) exhibit pronounced memory-bound characteristics during inference due to High Bandwidth Memory (HBM) bandwidth constraints. In this paper, we propose an L2 Cache-oriented asynchronous KV Cache prefetching method to break through the memory bandwidth bottleneck in LLM inference through computation-load overlap. By strategically scheduling idle memory bandwidth during active computation windows, our method proactively prefetches required KV Cache into GPU L2 cache, enabling high-speed L2 cache hits for subsequent accesses and effectively hiding HBM access latency within computational cycles. Extensive experiments on NVIDIA H20 GPUs demonstrate that the proposed method achieves 2.15× improvement in attention kernel efficiency and up to 1.97× end-to-end throughput enhancement, surpassing state-of-the-art baseline FlashAttention-3. Notably, our solution maintains orthogonality to existing optimization techniques and can be integrated with current inference frameworks, providing a scalable latency-hiding solution for next-generation LLM inference engines.

This paper proposed a neural networks-based adaptive control scheme for linear motors considering cogging force compensation. The cogging force is modeled and compensated for improving the tracking performance. An indirect parameter adaptive strategy is proposed to address the problem of parametric uncertainties. Compared with the direct adaptive strategy, this strategy can promote the parameter estimations to converge to the true values. In addition, radial basis neural networks are designed to estimate remaining system uncertainties, including unmodeled dynamics, model errors, and external disturbances. Comparative experiments are conducted on an iron-core permanent magnet linear synchronous motor platform. The experimental results show that the proposed control scheme can achieve excellent control performance.

The rapid adoption of AI and convenience offered by cloud services have resulted in the growing demands for GPUs in the cloud. Generally, GPUs are physically attached to host servers as PCIe devices. However, the fixed assembly combination of host servers and GPUs is extremely inefficient in resource utilization, upgrade, and maintenance. Due to these issues, the GPU disaggregation technique has been proposed to decouple GPUs from host servers. It aggregates GPUs into a pool and allocates GPU node(s) according to user demands. However, existing GPU disaggregation systems have flaws in software-hardware compatibility, disaggregation scope, and capacity. In this article, we present a new implementation of datacenter-scale GPU disaggregation, named DxPU. DxPU efficiently solves the above problems and can flexibly allocate as many GPU node(s) as users demand. To understand the performance overhead incurred by DxPU, we build up a performance model for AI specific workloads. With the guidance of modeling results, we develop a prototype system, which has been deployed into the datacenter of a leading cloud provider for a test run. We also conduct detailed experiments to evaluate the performance overhead caused by our system. The results show that the overhead of DxPU is less than 10%, compared with native GPU servers, in most of user scenarios.

The optimal operation of multi-reservoir system represents a complex nonlinear non-convex optimization problem. In this paper, an enhanced water cycle algorithm for optimization of multi-reservoir systems is proposed. The improvements refer to high-quality initial solutions obtained by the chaos-based method, balancing of exploration of streams using a dynamic adaptive parameter, and dynamic variation of sub-water system size using the fitness value of rivers. In order to verify the proposed improvements, the enhanced water cycle algorithm was applied to three typical benchmark functions and it was compared with water cycle algorithm and water cycle algorithm with evaporation rate. The results have shown that proposed algorithm provides performance improvement. The multi-reservoir system with complex constraints was selected for the case study. The results have shown that EWCA has higher ability to find a feasible solution in a narrow searching space than WCA and ER-WCA.