GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching

Sudipta Mondal; Susmita Dey Manasi; Kishor Kunal; S. Ramprasath; Sachin S. Sapatnekar

doi:10.1145/3489517.3530503

GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching

Sudipta Mondal, Susmita Dey Manasi, Kishor Kunal, S. Ramprasath, Sachin S. Sapatnekar

Electrical and Computer Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Scopus citations

Abstract

Graph neural networks (GNN) inferencing involves weighting vertex feature vectors, followed by aggregating weighted vectors over a vertex neighborhood. High and variable sparsity in the input vertex feature vectors, and high sparsity and power-law degree distributions in the adjacency matrix, can lead to (a) unbalanced loads and (b) inefficient random memory accesses. GNNIE ensures load-balancing by splitting features into blocks, proposing a flexible MAC architecture, and employing load (re)distribution. GNNIE's novel caching scheme bypasses the high costs of random DRAM accesses. GNNIE shows high speedups over CPUs/GPUs; it is faster and runs a broader range of GNNs than existing accelerators.

Original language	English (US)
Title of host publication	Proceedings of the 59th ACM/IEEE Design Automation Conference, DAC 2022
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	565-570
Number of pages	6
ISBN (Electronic)	9781450391429
DOIs	https://doi.org/10.1145/3489517.3530503
State	Published - Jul 10 2022
Event	59th ACM/IEEE Design Automation Conference, DAC 2022 - San Francisco, United States Duration: Jul 10 2022 → Jul 14 2022

Publication series

Name	Proceedings of the 59th ACM/IEEE Design Automation Conference

Conference

Conference	59th ACM/IEEE Design Automation Conference, DAC 2022
Country/Territory	United States
City	San Francisco
Period	7/10/22 → 7/14/22

Bibliographical note

Funding Information:
This work was supported in part by the Semiconductor Research Corporation (SRC).

Publisher Copyright:
© 2022 ACM.

Keywords

GNN
graph-specific caching
hardware accelerator
load balancing

Publisher link

10.1145/3489517.3530503

Find It

Find It at U of M Libraries

Cite this

Mondal, S., Manasi, S. D., Kunal, K., Ramprasath, S., & Sapatnekar, S. S. (2022). GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching. In Proceedings of the 59th ACM/IEEE Design Automation Conference, DAC 2022 (pp. 565-570). (Proceedings of the 59th ACM/IEEE Design Automation Conference). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1145/3489517.3530503

GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching. / Mondal, Sudipta; Manasi, Susmita Dey; Kunal, Kishor et al.
Proceedings of the 59th ACM/IEEE Design Automation Conference, DAC 2022. Institute of Electrical and Electronics Engineers Inc., 2022. p. 565-570 (Proceedings of the 59th ACM/IEEE Design Automation Conference).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Mondal, S, Manasi, SD, Kunal, K, Ramprasath, S & Sapatnekar, SS 2022, GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching. in Proceedings of the 59th ACM/IEEE Design Automation Conference, DAC 2022. Proceedings of the 59th ACM/IEEE Design Automation Conference, Institute of Electrical and Electronics Engineers Inc., pp. 565-570, 59th ACM/IEEE Design Automation Conference, DAC 2022, San Francisco, United States, 7/10/22. https://doi.org/10.1145/3489517.3530503

Mondal S, Manasi SD, Kunal K, Ramprasath S , Sapatnekar SS. GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching. In Proceedings of the 59th ACM/IEEE Design Automation Conference, DAC 2022. Institute of Electrical and Electronics Engineers Inc. 2022. p. 565-570. (Proceedings of the 59th ACM/IEEE Design Automation Conference). doi: 10.1145/3489517.3530503

@inproceedings{fac10bdfaf29456ebff3e625265225e2,

title = "GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching",

abstract = "Graph neural networks (GNN) inferencing involves weighting vertex feature vectors, followed by aggregating weighted vectors over a vertex neighborhood. High and variable sparsity in the input vertex feature vectors, and high sparsity and power-law degree distributions in the adjacency matrix, can lead to (a) unbalanced loads and (b) inefficient random memory accesses. GNNIE ensures load-balancing by splitting features into blocks, proposing a flexible MAC architecture, and employing load (re)distribution. GNNIE's novel caching scheme bypasses the high costs of random DRAM accesses. GNNIE shows high speedups over CPUs/GPUs; it is faster and runs a broader range of GNNs than existing accelerators.",

keywords = "GNN, graph-specific caching, hardware accelerator, load balancing",

author = "Sudipta Mondal and Manasi, {Susmita Dey} and Kishor Kunal and S. Ramprasath and Sapatnekar, {Sachin S.}",

note = "Funding Information: This work was supported in part by the Semiconductor Research Corporation (SRC). Publisher Copyright: {\textcopyright} 2022 ACM.; 59th ACM/IEEE Design Automation Conference, DAC 2022 ; Conference date: 10-07-2022 Through 14-07-2022",

year = "2022",

month = jul,

day = "10",

doi = "10.1145/3489517.3530503",

language = "English (US)",

series = "Proceedings of the 59th ACM/IEEE Design Automation Conference",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "565--570",

booktitle = "Proceedings of the 59th ACM/IEEE Design Automation Conference, DAC 2022",

}

TY - GEN

T1 - GNNIE

T2 - 59th ACM/IEEE Design Automation Conference, DAC 2022

AU - Mondal, Sudipta

AU - Manasi, Susmita Dey

AU - Kunal, Kishor

AU - Ramprasath, S.

AU - Sapatnekar, Sachin S.

PY - 2022/7/10

Y1 - 2022/7/10

N2 - Graph neural networks (GNN) inferencing involves weighting vertex feature vectors, followed by aggregating weighted vectors over a vertex neighborhood. High and variable sparsity in the input vertex feature vectors, and high sparsity and power-law degree distributions in the adjacency matrix, can lead to (a) unbalanced loads and (b) inefficient random memory accesses. GNNIE ensures load-balancing by splitting features into blocks, proposing a flexible MAC architecture, and employing load (re)distribution. GNNIE's novel caching scheme bypasses the high costs of random DRAM accesses. GNNIE shows high speedups over CPUs/GPUs; it is faster and runs a broader range of GNNs than existing accelerators.

AB - Graph neural networks (GNN) inferencing involves weighting vertex feature vectors, followed by aggregating weighted vectors over a vertex neighborhood. High and variable sparsity in the input vertex feature vectors, and high sparsity and power-law degree distributions in the adjacency matrix, can lead to (a) unbalanced loads and (b) inefficient random memory accesses. GNNIE ensures load-balancing by splitting features into blocks, proposing a flexible MAC architecture, and employing load (re)distribution. GNNIE's novel caching scheme bypasses the high costs of random DRAM accesses. GNNIE shows high speedups over CPUs/GPUs; it is faster and runs a broader range of GNNs than existing accelerators.

KW - GNN

KW - graph-specific caching

KW - hardware accelerator

KW - load balancing

UR - http://www.scopus.com/inward/record.url?scp=85137476146&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85137476146&partnerID=8YFLogxK

U2 - 10.1145/3489517.3530503

DO - 10.1145/3489517.3530503

M3 - Conference contribution

AN - SCOPUS:85137476146

T3 - Proceedings of the 59th ACM/IEEE Design Automation Conference

SP - 565

EP - 570

BT - Proceedings of the 59th ACM/IEEE Design Automation Conference, DAC 2022

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 10 July 2022 through 14 July 2022

ER -

GNNIE: GNN Inference Engine with Load-balancing and Graph-specific Caching

Abstract

Publication series

Conference

Bibliographical note

Keywords

Publisher link

Find It

Other files and links

Fingerprint

Cite this