arXiv subjects

Guanhua Li

Publications and source records attributed to Guanhua Li.

4 recordsLinked to original sources

arXiv2026

UBEP: Re-architecting Expert Parallelism Communication Library for Production Superpods

The deployment of Mixture-of-Experts (MoE) models on production high-bandwidth superpods, such as NVIDIA's NVL72/576 and Huawei's CloudMatrix384, introduces critical challenges beyond raw interconnect bandwidth. While these systems provide unified global address spaces and high-bandwidth fabrics, their full potential for sparse MoE communication is hindered by three fundamental bottlenecks: (1) Strict execution serialization imposed by coarse-grained Bulk Synchronous Parallel (BSP) orchestration of interdependent communication phases; (2) Prohibitive synchronization overhead that fails to scale alongside high interconnect bandwidth; and (3) Severe load imbalance resulting from distance-agnostic scheduling of irregular token traffic. To eliminate these bottlenecks, we introduce UBEP (Unified-Bus Expert Parallelism), a production-ready communication library that rethinks MoE's All-to-All primitives for modern superpod architectures. Through large scale experiments, UBEP reduces All-to-All latency by up to 52.4% and MoE inference Time Per Output Token (TPOT) by up to 11.1%.

cs.DC↗

arXiv2025

CAPSim: A Fast CPU Performance Simulator Using Attention-based Predictor

CPU simulators are vital for computer architecture research, primarily for estimating performance under different programs. This poses challenges for fast and accurate simulation of modern CPUs, especially in multi-core systems. Modern CPU peformance simulators such as GEM5 adopt the cycle-accurate and event-driven approach, which is timeconsuming to simulate the extensive microarchitectural behavior of a real benchmark running on out-of-order CPUs. Recently, machine leaning based approach has been proposed to improve simulation speed, but they are currently limited to estimating the cycles of basic blocks rather than the complete benchmark program. This paper introduces a novel ML-based CPU simulator named CAPSim, which uses an attention-based neural network performance predictor and instruction trace sampling method annotated with context. The attention mechanism effectively captures long-range influence within the instruction trace, emphasizing critical context information. This allows the model to improve performance prediction accuracy by focusing on important code instruction. CAPSim can predict the execution time of unseen benchmarks at a significantly fast speed compared with an accurate O3 simulator built with gem5. Our evaluation on a commercial Intel Xeon CPU demonstrates that CAPSim achieves a 2.2 - 8.3x speedup compared to using gem5 built simulator, which is superior to the cutting-edge deep learning approach

cs.PF↗

arXiv2017

Ultrasensitive biosensor based on Nd:YAG waveguide laser: Tumor cell and Dextrose solution

This work demonstrates the Nd:YAG waveguide laser as an efficient platform for the bio-sensing. The waveguide was fabricated in the Nd:YAG crystal by the cooperation of the ultrafast laser writing and ion irradiation. As the laser oscillation in the Nd:YAG waveguide is ultra-sensitivity to the external environment of the waveguide. Even a weak disturbance would induce a large variation of the output power of the laser. According to this feature, the Nd:YAG waveguide coated with Graphene and WSe2 layers is used as substrate for the microfluidic channel. When the microflow crosses the Nd:YAG waveguide, the laser oscillation in the waveguide is disturbed, and induces the fluctuation of the output laser. Through the analysis of the fluctuation, the concentration of the dextrose solution and the size of the tumor cell are distinguished

physics.app-ph↗

arXiv2010

A note on equilibrium Glauber and Kawasaki dynamics for permanental point processes

We construct two types of equilibrium dynamics of an infinite particle system in a locally compact metric space $X$ for which a permanental point process is a symmetrizing, and hence invariant measure. The Glauber dynamics is a birth-and-death process in $X$, while in the Kawasaki dynamics interacting particles randomly hop over $X$. In the case $X=\mathbb R^d$, we consider a diffusion approximation for the Kawasaki dynamics at the level of Dirichlet forms. This leads us to an equilibrium dynamics of interacting Brownian particles for which a permanental point process is a symmetrizing measure.

math.PR↗