CoDec: Prefix-Shared Decoding Kernel for LLMs
Year: 2026
Type: article
Abstract: Prefix-sharing among multiple prompts presents opportunities to combine the operations of the shared prefix, while attention computation in the decode stage, which becomes a critical bottleneck with increasing context lengths, is a memory-intensive process requiring heavy memory access on the key-value (KV) cache of the prefixes. Therefore, in this paper, we explore the potential of prefix-sharing in the attention computation of the decode stage. However, the tree structure of the prefix-sharing mechanism presents significant challenges for attention computation in efficiently processing shared KV cache access patterns while managing complex dependencies and balancing irregular workloads. To address the above challenges, we propose a dedicated attention kernel to co mbine the memory access of shared prefixes in the dec oding stage, namely CoDec. CoDec delivers two key innovations: a novel shared-prefix attention kernel that optimizes memory hierarchy and exploits both intra-block and inter-block parallelism, and a comprehensive workload balancing mechanism that efficiently estimates cost, divides tasks, and schedules execution. Experimental results show that CoDec achieves an average 1.9× speedup and 120.9× memory access reduction compared to the state-of-the-art FlashDecoding kernel regarding attention computation in the decode stage and up to 3.8× end-to-end time per output token speedup compared to vLLM.
DOI: 10.1145/3802028
Institutions: Nanjing University, Alibaba Group (China)
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7161581731