Uniconn: A Uniform High-Level Communication Library for Portable Multi-GPU Programming

Modern HPC and AI systems increasingly rely on multi-GPU clusters, where communication libraries such as MPI, NCCL/RCCL, and NVSHMEM enable data movement across GPUs. While these libraries are widely used in frameworks and solver packages, their distinct APIs, synchronization models, and integration...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Proceedings / IEEE International Conference on Cluster Computing S. 1 - 12
Hauptverfasser: Sagbili, Dogan, Ekmekcibasi, Sinan, Ibrahim, Khaled Z., Nguyen, Tan, Unat, Didem
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 02.09.2025
Schlagworte:
Benchmark testing
C++ languages
Codes
communication libraries
Complexity theory
GPU
Graphics processing units
Jacobian matrices
Libraries
MPI
Multi-GPUs
NCCL/RCCL
NVSHMEM
Performance evaluation
Programming
Switches
ISSN:2168-9253
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Modern HPC and AI systems increasingly rely on multi-GPU clusters, where communication libraries such as MPI, NCCL/RCCL, and NVSHMEM enable data movement across GPUs. While these libraries are widely used in frameworks and solver packages, their distinct APIs, synchronization models, and integration mechanisms introduce programming complexity and limit portability. Performance also varies across workloads and system architectures, making it difficult to achieve consistent efficiency. These issues present a significant obstacle to writing portable, high-performance code for large-scale GPU systems. We present Uniconn, a unified, portable high-level C++ communication library that supports both point-to-point and collective operations across GPU clusters. Uniconn enables seamless switching between backends and APIs (host or device) with minimal or no changes to application code. We describe its design and core constructs, and evaluate its performance using network benchmarks, a Jacobi solver, and a Conjugate Gradient solver. Across three supercomputers, we compare Uniconn's overhead against CUDA/ROCm-aware MPI, NCCL/RCCL, and NVSHMEM on up to 64 GPUs. In most cases, Uniconn incurs negligible overhead, typically under 1 % for the Jacobi solver and under 2% for the Conjugate Gradient solver.
ISSN:2168-9253
DOI:10.1109/CLUSTER59342.2025.11186498