SemBridge creates clear communication plans between different AI runtimes
SemBridge: Compiling Consumer Observations into Cross-Stack Communication Plans
Distributed, Parallel, and Cluster Computing
Summary
When different AI systems work together, they need to share information clearly, but sometimes they speak different technical languages. The researchers developed SemBridge, a method that builds clear 'contracts' explaining what data one system needs to see and how it should be delivered. This helps different parts of AI software coordinate without confusion or errors, even when using different hardware setups like NVIDIA CUDA or Huawei CANN. Their tests showed SemBridge reduces unnecessary data transfer and speeds up processing across various AI workloads.
distributed tensor systemscollective communicationruntimecommunication planCUDANCCLCANNHCCLconsumer observationcapacity spillover
Authors
Genlang Chen, Junyi Zhu, Yuanshan Lin
Abstract
Distributed-tensor systems specify where values reside, while collective systems optimize how requested operations execute. At a boundary between vendor runtimes that cannot share a native communicator, neither abstraction states what a remote consumer must observe. SemBridge fills this gap by compiling graph and runtime facts into a typed contract for the consumer-visible result and its delivery obligations. The contract captures provenance, substitutability, completion, authority, demand, and native-domain locality. A deterministic lowerer constructs backend-neutral communication plans, and a symbolic checker validates each plan before execution across CUDA/NCCL and CANN/HCCL. An independent layout-only planner handles all 72 structural transitions but establishes only 54 complete obligations; a byte-only minimizer proposes 40 semantically invalid candidates, all rejected by SemBridge. On nine real edges, SemBridge produces distinct observation-aware plans that reduce startups on all nine and payload bytes on the three result edges. A live CUDA/CANN run derives and executes full-logit reconstruction, source projection, and owner-token delivery from log-probability, token-only, and owner-scoped requests. On a measured two-host 1-GbE capacity-spillover deployment, source projection cuts result traffic by more than 99.97% and increases throughput by 8.92-80.20% across Dense, MoE, and MiniMax workloads. All 18 MiniMax restart pairs at concurrency 1, 8, and 16 favor source projection. A Qwen3-14B MLP slice additionally verifies bitwise activation-shard delivery and HCCL completion of row-parallel partials. These results establish consumer observation as a semantic layer between placement and collective execution.