Skip to content

Commit

Permalink
Merge pull request #2 from hpcde/news-hpdc23-spmv-paper
Browse files Browse the repository at this point in the history
add HPDC23 SpMV paper news

We have published a new paper: http://doi.org/10.1145/3588195.3593002
  • Loading branch information
genshen committed May 10, 2023
2 parents a6ef1a6 + e19f203 commit 305b66f
Show file tree
Hide file tree
Showing 3 changed files with 59 additions and 0 deletions.
57 changes: 57 additions & 0 deletions content/zh/post/2023-hpdc-spmv-paper/index.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,57 @@
---
title: "喜讯:胡长军教授团队论文被高性能计算领域重要会议 HPDC 2023 接收"
date: 2023-05-10T21:01:00+08:00
# authors: GPU team
draft: false
authors:
- admin
---

日前,课题组胡长军教授团队储根深博士等人完成的论文 “Efficient Algorithm Design of Optimizing SpMV on GPU” 被 HPDC 2023 接收。

[高性能并行和分布式计算国际研讨会](https://hpdc.org)(International Symposium on High-Performance Parallel and Distributed Computing,HPDC)旨在介绍有关高性能计算并行和分布式系统设计、实现、评估和使用的最新研究,是大型计算机研究中久负盛名的会议之一([CS-Ranking](https://csrankings.org)上列出的高性能计算领域三大顶会之一)。
HPDC 2023 将于6月中旬在美国弗罗里达州奥兰多市召开,该会议从 100 多篇投稿论文中,最终评审出 22 篇高质量论文录用。

<!--more-->

论文 “Efficient Algorithm Design of Optimizing SpMV on GPU” 面向 GPU 架构上的稀疏向量乘(SpMV)这一重要 kernel,开展了新的算法设计与性能优化,提出了两个新的高效 SpMV 并行算法并通过自适应的方式进行组合,将 SpMV 的性能进行了大量提升,突破了前述算法的性能纪录。

一直以来,稀疏矩阵向量在科学计算中发挥着重要作用,自 GPU 出现以来的数十年里,高性能计算的科研人员和高性能计算工程师们一直在开展 SpMV 的性能调优,以不断逼近硬件性能极限。
本文提出的SpMV 的算法,对 GPU 上的负载均衡、访存性能、高速片上缓存和调度开展了深入研究。

![line-enhance 算法](/images/news/2023-spmv-hpdc23/spmv-line-enhance.svg)
*本论文提出的新算法示意图 (其一的 line-enhance 算法)*

最终实现的算法相对于目前的 SOTA (state of the art) 算法,
性能分别有了424%(对比 CSR-Vector 算法), 741% (对比 CSR-Adaptive 算法), 49%(对比 HOLA 算法), 46%(对比 cuSparse 算法), 72%(对比 merge-based SpMV 算法)的提升。
该成果得到审稿人的充分肯定,并评价说 “Performance improvement compared to cuSparse and ROCm are impressive.”(注:cuSparse 与 ROCm 分别是 NVIDIA 和 AMD GPU的系统库)。

![](/images/news/2023-spmv-hpdc23/spmv-vs-cub-speedup.svg)
*图:本论文提出的新算法性能与CUDA 生态的 cub 库的性能对比 (纵坐标为加速比,横坐标为稀疏矩阵算例,按矩阵非零元素排列,大于1表述我们算法性能更优)*

团队一直以来面向国产超算开展高性能计算软件、算法库的开发与性能优化工作,目前该系列算法已经适配曙光超算的 DCU 架构,期待后续该系列算法可以在国产超算生态中发挥重要作用。

---
论文信息:
```bib
# For github user: you can also click the link in this repository landing page at the right sidebar, with the label "Cite this repository."
@inproceedings{chu_spmv_gpu:_2023,
title = {Efficient Algorithm Design of Optimizing SpMV on GPU},
isbn = {979-8-4007-0155-9/23/06},
url = {http://doi.org/10.1145/3588195.3593002},
doi = {10.1145/3588195.3593002},
language = {en},
urldate = {2023-6-20},
booktitle = {Proceedings of the 32nd International Symposium on High-Performance Parallel and Distributed Computing (HPDC '23), June 16--23, 2023, Orlando, FL, USA},
publisher = {ACM Press},
author = {Chu, Genshen and He, Yuanjie and Dong, Lingyu and Ding, Zhezhao and Chen, Dandan and Bai, He and Wang, Xuesong and Hu, Changjun},
year = {2023},
numpages = {14},
series = {HPDC '23},
address = {Orlando, Florida},
location = {Orlando, FL, USA},
pages = {1--14},
}
```

论文链接:[http://doi.org/10.1145/3588195.3593002](http://doi.org/10.1145/3588195.3593002)
1 change: 1 addition & 0 deletions static/images/news/2023-spmv-hpdc23/spmv-line-enhance.svg
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.

0 comments on commit 305b66f

Please sign in to comment.