ZipServ: Fast and Memory-Efficient LLM Inference with Hardware-Aware Lossless Compression

TL;DR

ZipServ is a novel lossless compression framework for LLM inference that significantly reduces model size and accelerates GPU-based inference by co-designing compression and computation kernels for hardware efficiency.

Contribution

It introduces TCA-TBE encoding and a fused decompression-GEMM kernel, enabling efficient, lossless, and hardware-aware LLM inference acceleration.

Findings

Model size reduced by up to 30%

Achieves up to 2.21x kernel speedup

Speeds up end-to-end inference by 1.22x

Abstract

Lossless model compression holds tremendous promise for alleviating the memory and bandwidth bottlenecks in bit-exact Large Language Model (LLM) serving. However, existing approaches often result in substantial inference slowdowns due to fundamental design mismatches with GPU architectures: at the kernel level, variable-length bitstreams produced by traditional entropy codecs break SIMT parallelism; at the system level, decoupled pipelines lead to redundant memory traffic. We present ZipServ, a lossless compression framework co-designed for efficient LLM inference. ZipServ introduces Tensor-Core-Aware Triple Bitmap Encoding (TCA-TBE), a novel fixed-length format that enables constant-time, parallel decoding, together with a fused decompression-GEMM (ZipGEMM) kernel that decompresses weights on-the-fly directly into Tensor Core registers. This "load-compressed, compute-decompressed" design eliminates intermediate buffers and maximizes compute intensity. Experiments show that ZipServ reduces the model size by up to 30%, achieves up to 2.21x kernel-level speedup over NVIDIA's cuBLAS, and expedites end-to-end inference by an average of 1.22x over vLLM. ZipServ is the first lossless compression system that provides both storage savings and substantial acceleration for LLM inference on GPUs.