BNU-SYS Lab

BNU-SYS Lab

Research on computer architecture, parallel and high-performance computing, code analysis/optimization, and LLM inference optimization.

Posts by Tags

APPT

高建花参加 APPT 2026 并作论文报告

less than 1 minute read

Published:

课题组高建花老师在比利时布鲁塞尔举行的 APPT 2026 上作题为 ARROW 的学术报告，提出面向 SpMV 的自适应行重组方法，在张量核上实现 warp 级负载均衡，性能优于 cuSPARSE、DASP 与 Spaden 等主流方法。

Jianhua Gao Presented at APPT 2026

1 minute read

Published:

Jianhua Gao from the group presented the ARROW study at APPT 2026 in Brussels, Belgium, proposing an adaptive row reorganization method for SpMV that achieves warp-level load balancing on Tensor Cores and outperforms mainstream methods such as cuSPARSE, DASP, and Spaden.

ASC

Ascend

BNU-SYS

葛丹颖参加 Euro-Par 2026 并作论文报告

less than 1 minute read

Published:

葛丹颖在第 32 届欧洲并行与分布式计算会议（Euro-Par 2026）上汇报了面向检索增强生成（RAG）的推理加速工作 ApproxRAG，通过协同优化磁盘、主机内存与设备内存的数据流动，实现最高 3.66 倍推理加速。

Danying Ge Presented at Euro-Par 2026

1 minute read

Published:

At the 32nd International European Conference on Parallel and Distributed Computing (Euro-Par 2026), Danying Ge presented ApproxRAG, an inference acceleration work for retrieval-augmented generation (RAG) that jointly optimizes data movement among disk, host memory, and device memory to achieve up to a 3.66× inference speedup.

高建花参加 APPT 2026 并作论文报告

less than 1 minute read

Published:

课题组高建花老师在比利时布鲁塞尔举行的 APPT 2026 上作题为 ARROW 的学术报告，提出面向 SpMV 的自适应行重组方法，在张量核上实现 warp 级负载均衡，性能优于 cuSPARSE、DASP 与 Spaden 等主流方法。

Jianhua Gao Presented at APPT 2026

1 minute read

Published:

Jianhua Gao from the group presented the ARROW study at APPT 2026 in Brussels, Belgium, proposing an adaptive row reorganization method for SpMV that achieves warp-level load balancing on Tensor Cores and outperforms mainstream methods such as cuSPARSE, DASP, and Spaden.

Challenge Track

Euro-Par

葛丹颖参加 Euro-Par 2026 并作论文报告

less than 1 minute read

Published:

葛丹颖在第 32 届欧洲并行与分布式计算会议（Euro-Par 2026）上汇报了面向检索增强生成（RAG）的推理加速工作 ApproxRAG，通过协同优化磁盘、主机内存与设备内存的数据流动，实现最高 3.66 倍推理加速。

Danying Ge Presented at Euro-Par 2026

1 minute read

Published:

At the 32nd International European Conference on Parallel and Distributed Computing (Euro-Par 2026), Danying Ge presented ApproxRAG, an inference acceleration work for retrieval-augmented generation (RAG) that jointly optimizes data movement among disk, host memory, and device memory to achieve up to a 3.66× inference speedup.

Group Competition Award

Huawei ICT Competition

Kunpeng

Qwen

RAG

葛丹颖参加 Euro-Par 2026 并作论文报告

less than 1 minute read

Published:

葛丹颖在第 32 届欧洲并行与分布式计算会议（Euro-Par 2026）上汇报了面向检索增强生成（RAG）的推理加速工作 ApproxRAG，通过协同优化磁盘、主机内存与设备内存的数据流动，实现最高 3.66 倍推理加速。

Danying Ge Presented at Euro-Par 2026

1 minute read

Published:

At the 32nd International European Conference on Parallel and Distributed Computing (Euro-Par 2026), Danying Ge presented ApproxRAG, an inference acceleration work for retrieval-augmented generation (RAG) that jointly optimizes data movement among disk, host memory, and device memory to achieve up to a 3.66× inference speedup.

SpMV

高建花参加 APPT 2026 并作论文报告

less than 1 minute read

Published:

课题组高建花老师在比利时布鲁塞尔举行的 APPT 2026 上作题为 ARROW 的学术报告，提出面向 SpMV 的自适应行重组方法，在张量核上实现 warp 级负载均衡，性能优于 cuSPARSE、DASP 与 Spaden 等主流方法。

Jianhua Gao Presented at APPT 2026

1 minute read

Published:

Jianhua Gao from the group presented the ARROW study at APPT 2026 in Brussels, Belgium, proposing an adaptive row reorganization method for SpMV that achieves warp-level load balancing on Tensor Cores and outperforms mainstream methods such as cuSPARSE, DASP, and Spaden.

Student Supercomputer Challenge

Tensor Core

Jianhua Gao Presented at APPT 2026

1 minute read

Published:

Jianhua Gao from the group presented the ARROW study at APPT 2026 in Brussels, Belgium, proposing an adaptive row reorganization method for SpMV that achieves warp-level load balancing on Tensor Cores and outperforms mainstream methods such as cuSPARSE, DASP, and Spaden.

XianDaoBei

academic conference

Danying Ge Presented at Euro-Par 2026

1 minute read

Published:

At the 32nd International European Conference on Parallel and Distributed Computing (Euro-Par 2026), Danying Ge presented ApproxRAG, an inference acceleration work for retrieval-augmented generation (RAG) that jointly optimizes data movement among disk, host memory, and device memory to achieve up to a 3.66× inference speedup.

Jianhua Gao Presented at APPT 2026

1 minute read

Published:

Jianhua Gao from the group presented the ARROW study at APPT 2026 in Brussels, Belgium, proposing an adaptive row reorganization method for SpMV that achieves warp-level load balancing on Tensor Cores and outperforms mainstream methods such as cuSPARSE, DASP, and Spaden.

award

category1

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

category2

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

computer system capability

cool posts

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

high-performance computing

Jianhua Gao Presented at APPT 2026

1 minute read

Published:

Jianhua Gao from the group presented the ARROW study at APPT 2026 in Brussels, Belgium, proposing an adaptive row reorganization method for SpMV that achieves warp-level load balancing on Tensor Cores and outperforms mainstream methods such as cuSPARSE, DASP, and Spaden.

large-model inference

Danying Ge Presented at Euro-Par 2026

1 minute read

Published:

At the 32nd International European Conference on Parallel and Distributed Computing (Euro-Par 2026), Danying Ge presented ApproxRAG, an inference acceleration work for retrieval-augmented generation (RAG) that jointly optimizes data movement among disk, host memory, and device memory to achieve up to a 3.66× inference speedup.

research planning

supercomputing

team news

year-end review

世界大学生超级计算机竞赛

先导杯

千问

华为ICT大赛

团队新闻

大模型推理

葛丹颖参加 Euro-Par 2026 并作论文报告

less than 1 minute read

Published:

葛丹颖在第 32 届欧洲并行与分布式计算会议（Euro-Par 2026）上汇报了面向检索增强生成（RAG）的推理加速工作 ApproxRAG，通过协同优化磁盘、主机内存与设备内存的数据流动，实现最高 3.66 倍推理加速。

学术会议

葛丹颖参加 Euro-Par 2026 并作论文报告

less than 1 minute read

Published:

葛丹颖在第 32 届欧洲并行与分布式计算会议（Euro-Par 2026）上汇报了面向检索增强生成（RAG）的推理加速工作 ApproxRAG，通过协同优化磁盘、主机内存与设备内存的数据流动，实现最高 3.66 倍推理加速。

高建花参加 APPT 2026 并作论文报告

less than 1 minute read

Published:

课题组高建花老师在比利时布鲁塞尔举行的 APPT 2026 上作题为 ARROW 的学术报告，提出面向 SpMV 的自适应行重组方法，在张量核上实现 warp 级负载均衡，性能优于 cuSPARSE、DASP 与 Spaden 等主流方法。

年终总结

张量核

高建花参加 APPT 2026 并作论文报告

less than 1 minute read

Published:

课题组高建花老师在比利时布鲁塞尔举行的 APPT 2026 上作题为 ARROW 的学术报告，提出面向 SpMV 的自适应行重组方法，在张量核上实现 warp 级负载均衡，性能优于 cuSPARSE、DASP 与 Spaden 等主流方法。

挑战赛

昇腾

科研规划

竞赛获奖

计算机系统能力大赛

超智融合

超级团队奖

高性能计算

高建花参加 APPT 2026 并作论文报告

less than 1 minute read

Published:

课题组高建花老师在比利时布鲁塞尔举行的 APPT 2026 上作题为 ARROW 的学术报告，提出面向 SpMV 的自适应行重组方法，在张量核上实现 warp 级负载均衡，性能优于 cuSPARSE、DASP 与 Spaden 等主流方法。

鲲鹏