Gravel fine-grain GPU-initiated network messages

Distributed systems incorporate GPUs because they provide massive parallelism in an energy-efficient manner. Unfortunately, existing programming models make it difficult to route a GPU-initiated network message. The traditional coprocessor model forces programmers to manually route messages through...

Full description

Saved in:
Bibliographic Details
Published in:International Conference for High Performance Computing, Networking, Storage and Analysis (Online) pp. 1 - 12
Main Authors: Orr, Marc S., Che, Shuai, Beckmann, Bradford M., Oskin, Mark, Reinhardt, Steven K., Wood, David A.
Format: Conference Proceeding
Language:English
Published: New York, NY, USA ACM 12.11.2017
Series:ACM Conferences
Subjects:
Computational modeling
Computing methodologies > Parallel computing methodologies > Parallel algorithms > Massively parallel algorithms
fine-grain communication
graphics processing unit (GPU)
Graphics processing units
message aggregation
Network topology
partitioned global address space (PGAS)
Programming
Prototypes
Semantics
Writing
graphics processing unit (GPU)
partitioned global address space (PGAS)
message aggregation
fine-grain communication
ISBN:9781450351140, 145035114X
ISSN:2167-4337
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Distributed systems incorporate GPUs because they provide massive parallelism in an energy-efficient manner. Unfortunately, existing programming models make it difficult to route a GPU-initiated network message. The traditional coprocessor model forces programmers to manually route messages through the host CPU. Other models allow GPU-initiated communication, but are inefficient for small messages. To enable fine-grain PGAS-style communication between threads executing on different GPUs, we introduce Gravel. GPU-initiated messages are offloaded through a GPU-efficient concurrent queue to an aggregator (implemented with CPU threads), which combines messages targeting to the same destination. Gravel leverages diverged work-group-level semantics to amortize synchronization across the GPU's data-parallel lanes. Using Gravel, we can distribute six applications, each with frequent small messages, across a cluster of eight GPU-accelerated nodes. Compared to one node, these applications run 5.3x faster, on average. Furthermore, we show Gravel is more programmable and usually performs better than prior GPU networking models.
ISBN:9781450351140
145035114X
ISSN:2167-4337
DOI:10.1145/3126908.3126914