NCCLZ: Compression-Enabled GPU Collectives with Decoupled Quantization and Entropy Coding
Open Access
- 13 July 2026
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 577-578
Abstract
Large multi-node GPU jobs often become limited by the number of bytes injected into the inter-node network path. Compression can reduce this traffic, but existing compression-assisted collective libraries either sit outside NCCL’s GPU-centric execution model or mainly rely on quantization alone. We present NCCLZ, a compression-enabled extension to NCCL that decouples lossy quantization from lossless entropy coding. NCCLZ performs approximation once before communication by converting tensors into integer symbols, and then applies lossless byte reduction inside NCCL’s communication pipeline. This placement keeps numerical approximation explicit, avoids repeated lossy recompression during reductions, and preserves NCCL’s native pipelined execution model. NCCLZ combines a runtime selector with slot-aware framing and pipelined staging to choose among raw transmission, fixed-length packing, and GPU Huffman coding. Initial results on Polaris show up to 9.65× compression ratio and up to 4.79× higher end-to-end AllReduce throughput than baseline NCCL on representative scientific and training tensors.Keywords
All Related Versions
This publication has 4 references indexed in Scilit:
- ghZCCL: Advancing GPU-aware Collective Communications with Homomorphic Compression2025
- gZCCL: Compression-Accelerated Collective Communication Framework for GPU Clusters2024
- cuSZp: An Ultra-fast GPU Error-bounded Lossy Compression Framework with Optimized End-to-End Performance2023
- Lightweight Huffman Coding for Efficient GPU Compression2023