Similar Items: Eliminating Hidden Serialization in Multi-Node Megakernel Communication
- SuperPaymaster: Eliminating Centralized Signer Authority via Asset-Oriented Abstraction to Reconcile Usability and Decentralization in Account Abstraction
- MoE-Hub: Taming Software Complexity for Seamless MoE Overlap with Hardware-Accelerated Communication on Multi-GPU Systems
- Communication Offloading on SmartNIC DPUs: A Quantitative Approach
- SparseRL-Sync: Lossless Weight Synchronization with ~100x Less Communication
- ZipCCL: Efficient Lossless Data Compression of Communication Collectives for Accelerating LLM Training
- Relay Buffer Independent Communication over Pooled HBM for Efficient MoE Inference on Ascend