arrow
返回

Disaggregated Design for GPU-Based Volumetric Data Structures

delete2026-01-01
delete1
PRE
AI
M
Massimiliano Meneghin
A
Ahmed H. Mahmoud *
DOI:10.1007/978-3-031-99872-0_11delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
体数据结构通常优先考虑数据局部性,专注于高效的内存访问模式。这种单一关注可能忽视其他关键性能因素,如占用率、通信和核融合。我们引入了一种新颖的解耦设计,重新平衡了局部性与这些目标之间的权衡——减少分布式内存架构上的通信开销,缓解复杂边界条件下的寄存器压力,并实现核融合。我们对单节点多GPU格子玻尔兹曼方法(LBM)求解器进行了全面分析其优势。我们的评估涵盖了密集、块稀疏和多分辨率离散化,展示了我们设计的灵活性和效率。借助这种方法,我们在现有最佳解决方案上实现了高达3倍的加速。
Keyword:
Data layout
Parallel
Simulation
GPU
LBM

期刊

E
EURO-PAR 2025: PARALLEL PROCESSING, PT III
IF:
0
论文数:
21
被引数:
0

机构

M
massachusetts institute of technology (mit)
学者数:
1.4K
论文数: 622
被引数: 0
引用论文

引用论文

Straight velocity boundaries in the lattice Boltzmann method
err2008-05-13
err0
PREAI
errJonas Latt; Bastien Chopard; Orestis Malaspinas; Michel Deville; Andreas Michler
err分享
err收藏
Boundaries in lattice gas flows
err1991-01-01
err0
PREAI
errPaul Lavallée; Jean Pierre Boon; Alain Noullez
err分享
err收藏
Cache-Oblivious Algorithms
err2012-01-01
err0
PREAI
errMatteo Frigo; Charles E. Leiserson; Harald Prokop; Sridhar Ramachandran
err分享
err收藏
A Massively Parallel and Scalable Multi-GPU Material Point Method一种大规模并行可扩展的多GPU材料点方法
err2020-08-12
err29
errOAAI
errWang, Xinlei; Qiu, Yuxing; Slattery, Stuart R.; Fang, Yu; Li, Minchen; Zhu, Song-Chun; Zhu, Yixin; Tang, Min; Manocha, Dinesh; Jiang, Chenfanfu
err分享
err收藏
LogP: towards a realistic model of parallel computation
err1993-07-01
err0
errOAAI
errDavid Culler; Richard Karp; David Patterson; Abhijit Sahay; Klaus Erik Schauser; Eunice Santos; Ramesh Subramonian; Thorsten von Eicken
err分享
err收藏
Cross-platform programming model for many-core lattice Boltzmann simulations
err2021-04-29
err0
errOAAI
errJonas Latt; Christophe Coreixas; Joël Beny
err分享
err收藏
Comparison of different propagation steps for lattice Boltzmann methods
err2013-03-01
err0
errOAAI
errMarkus Wittmann; Thomas Zeiser; Georg Hager; Gerhard Wellein
err分享
err收藏
学者 查看更多内容