一句话：TCCL 通过性能剖析与路径搜索，为 PCIe GPU 集群自动发现最优集合通信路径，并改造成 NCCL 运行时加以利用，最高提速 2.07 倍。
深度学习训练依赖 GPU 通过集合通信原语协作，但已有研究多集中于配备专用互连（如 DGX/NVLink）的系统，主要经由 PCIe 通信的 GPU 集群则关注不足——这类集群的拓扑与带宽异构性使得固定通信路径往往并非最优。本文提出专为 PCIe GPU 集群设计的集合通信库 TCCL，包含三个组件：用于测量多传输性能的 profiler、用于发现最优通信路径的 pathfinder，以及改造后的 NCCL 运行时以实际使用所发现的路径。工作聚焦于 AllReduce、AllGather 等深度学习常用操作所适用的环状（ring-based）集合通信算法。在三个不同的 PCIe GPU 集群上，TCCL 相对当前最先进的通信库 NCCL 与 MSCCL 最高取得 2.07 倍加速；作者还用多种并行策略组合的深度学习训练负载对 TCCL 进行了评估。
Exploiting parallelism to train deep learning models requires GPUs to cooperate through collective communication primitives. While systems like DGX, equipped with proprietary interconnects, have been extensively studied, the systems where GPUs mainly communicate through PCIe have received limited attention. This paper introduces TCCL, a collective communication library designed explicitly for such systems. TCCL has three components: a profiler for multi-transfer performance measurement, a pathfinder to discover optimal communication paths, and a modified runtime of NCCL to utilize the identified paths. The focus is on ring-based collective communication algorithms that apply to popular communication operations in deep learning, such as AllReduce and AllGather. The evaluation results of TCCL on three different PCIe-dependent GPU clusters show that TCCL outperforms (up to ×2.07) the state-of-the-art communication libraries, NCCL and MSCCL. We also evaluate TCCL with DL training workloads with various combinations of parallelism types.