会议日程

第30届中国计算机系统研讨会将于5月29日–5月31日在中国·西安举行。

5月29日

时间	安排
15:00-22:00	会议报到（希尔顿酒店签到）

5月30日

时间	安排
09:00–09:10	开场仪式
09:10–09:50	特邀报告一主持人：徐尔茨，上海交通大学嘉宾：中科院计算所，包云岗主题：关于如何在中国做出图灵奖级别科研工作的探讨
09:50–10:30	特邀报告二主持人：徐尔茨，上海交通大学嘉宾：阿里巴巴通义大模型事业部，翟恩南主题：算力变现的最后一公里：通义千问 Infra 的系统创新之路
10:30–10:50	上午茶歇（20分钟）
10:50–12:05	Session 1: ML Systems 1 主持人：章明星，清华大学 Sharpen the Spec, Cut the Code: A Case for Generative File System with SYSSPEC Dynamic Scheduling for AI Accelerators via TISA Efficient Temporal Graph Network Training via Unified Redundancy Elimination RobustRL: Role-based Fault Tolerance System for RL Post-Training AUM: Unleashing the Efficiency Potential of Shared Processors
12:05–13:30	午餐
13:30–15:00	Session 2: Storage & Database Systems 主持人：段卓辉，华中科技大学 Learned Query Optimizer in Alibaba MaxCompute: Challenges, Analysis, and Solutions CoPilotIO: CPU as a Co-pilot for GPU I/O to Free GPU Compute Nemo: A Low-Write-Amplification Cache for Tiny Objects on Log-Structured Flash Devices Wiseswap: Elastic Datacenter Network-Aware Disaggregated Memory for Multi-Tenant Cloud CPU-Oblivious Offloading of Failure-Atomic Transactions for Disaggregated Memory Merlin: An Efficient Adaptive Cache Eviction Algorithm via Fine-Grained Characterization
15:00–15:20	下午茶歇（20分钟）
15:20–16:20	Session 3: Demo Presentation 主持人：岳恒山，吉林大学 OpenSUN: An Open Platform for Scale-Up Network Systems Demonstration of FM-Agent: LLM-Based Automated Reasoning for Large-Scale System Code Wolvrix：面向芯片架构研究的 RTL 编译基础设施 OpenFlash: 面向敏捷设计与验证的开放 SSD 平台
16:20–17:35	Session 4: Distributed & Cloud Systems 主持人：董明凯，上海交通大学 Cohet: A CXL-Driven Coherent Heterogeneous Computing Framework Themis: Detecting Distributed Concurrency Bugs through RPC-Driven Race-Directed Test Generation and Fuzzing Nezha: A Key-Value Separated Distributed Store with Optimized Raft Integration NanoCP: Request-Level Dynamic Context Parallelism for MoE Serving Characterizing Trust Boundary Vulnerabilities in TEE Container Systems
17:35–18:05	产业论坛主持人：张余豪，天津大学蚂蚁，郑达，大语言模型时代的 AI 基础设施演化：从系统协同设计到自主优化华为，付明，Web 内核在 AI 时代的机遇和挑战腾讯，陈旭升，Hyinfra 实践：Solid 底座驱动的模型快速迭代与线上极致性价比
18:30–20:00	晚宴
20:30	执委会闭门会议

5月31日

时间	安排
09:00–09:10	颁奖环节
09:10–09:50	特邀报告一主持人：廖晓坚嘉宾：字节跳动，肖文聪主题：聚焦大模型训练基础设施：从优化单一任务到构建训练操作系统
09:50–10:30	特邀报告二主持人：廖晓坚嘉宾：华为，郭寒军主题：异构融合操作系统：Agent 负载长稳可靠运行和高效资源底座
10:30–10:50	上午茶歇（20分钟）
10:50–12:05	Session 5: ML Systems 2 主持人：赵涵，上海交通大学 LayerScope: Predictive Cross-Layer Scheduling for Efficient Multi-Batch MoE Inference on Legacy Servers KTransformers-FineTune: Beyond Offload with Layout-Aware and Schedule-Optimized Heterogeneous MoE Fine-Tuning MLX: Multi-Layer Execution for Structured LLM Workload Acceleration on Spatial Architectures Evaluation and Optimization of Communication Compression for 10K-GPU LLM Training SpecGMRES: Mixed Precision Speculative Execution of GMRES Iterative Algorithm
12:05–13:30	午餐
13:30–14:45	Session 6: Computer Architecture 主持人：汤战勇，西北大学 An Efficient and Scalable Hardware Architecture for Number Theoretic Transform on FPGA with Design Automation Peregrine: Accelerating TFHE Bootstrapping on GPUs via Multi-Level External Product Co-Design COMET: A Cooperative Scheduling Framework for Concurrent PIM/CPU Execution on Mobile Devices Optimizing Spatial Data Structure with Near-Cache Acceleration by Exploiting Physical Locality Revisiting Global Value Prediction: A Resurgent Complement to Local Predictors
14:45–15:00	中场休息（15分钟）
15:00–16:30	Session 7: Embedded & Interdisciplinary Systems 主持人：孟宪佳，西北大学 Graph.hls: A Compiler Framework for Composable Graph Accelerator Design OptiFX: Automatic Optimization for Convolutional Neural Networks with Aggressive Operator Fusion on GPUs Characterizing Matrix Multiplication Units across General Parallel Patterns in Scientific Computing AccShield: Operator-Aware Robust Scheduling for Fault-Tolerant AI Accelerators Technical Report: Uncovering and Mitigating Linux MGLRU's Generational Imbalance in Android at Scale Lifting Optimized Binaries to Canonical Compiler IR via Structure-Aware Retrieval and Iterative Verification
16:30	闭幕致辞暨会议结束