StageFrontier: Synchronization-Aware Stage Accounting for Distributed ML Training

Collection of recent systems and ML infrastructure papers (June 2026). Key contributions: StageFrontier — a low-overhead, always-on synchronization-aware accounting signal for distributed ML that pinpoints which stage and rank first expose group-visible delays (works with PyTorch, Gloo, NCCL). Terastal — layer-variant design and scheduling to reduce deadline misses on heterogeneous DNN accelerators. Multi-GPU 3D FDTD study — communication strategies showing direct GPU-to-GPU peer exchange dominates. PCCL — process-group-aware, scalable synthesizer for topology-aware collective algorithms. A治理: