STI Publications - View Publication Form #17021

Back to Search Results Print
Publication Information
Title 24 MemHC: An Optimized GPU Memory Management Framework for Accelerating Many-body Correlation
Abstract The many-body correlation function is a fundamental computation kernel in modern physics computing applications, e.g., Hadron Contractions in Lattice quantum chromodynamics (QCD). This kernel is both computation and memory intensive, involving a series of tensor contractions, and thus usually runs on accelerators like GPUs. Existing optimizations on many-body correlation mainly focus on individual tensor contractions (e.g., cuBLAS libraries and others). In contrast, this work discovers a new optimization dimension for many-body correlation by exploring the optimization opportunities among tensor contractions. More specifically, it targets general GPU architectures (both NVIDIA and AMD) and optimizes many-body correlation¿s memory management by exploiting a set of memory allocation and communication redundancy elimination opportunities: first, GPU memory allocation redundancy: the intermediate output frequently occurs as input in the subsequent calculations; second, CPU-GPU communicatio
Author(s) Qihan Wang, Zhen Peng, Bin Ren, Jie Chen, Robert Edwards
Publication Date June 2022
Document Type Journal Article
Primary Institution Thomas Jefferson National Accelerator Facility, Newport News
Affiliation Comp Sci&Tech (CST) Div / Scientific Computing / Scientific Computing
Funding Source Nuclear Physics (NP)
Proprietary? No
This publication conveys Technical Science Results
Document Numbers
JLAB Number: JLAB-CST-22-3602 OSTI Number: 1867362
LANL Number: Other Number: DOE/OR/23177-5487
Associated with an experiment No
Associated with EIC No
Supported by Jefferson Lab LDRD Funding No
Journal Article
Journal Name ACM Transactions on Architecture and Code Optimization
Refereed Yes
Volume 19
Issue 2
Page(s) 24
Attachments/Datasets/DOI Link
Document(s)
3506705.pdf (STI Document)
3506705.pdf (Accepted Manuscript)
DOI Link
Dataset(s) (none)
Back to Search Results Print