MegaOps: A Kernel Overlapping Framework with PyTorch-like Modularity and Megakernel Performance
cs.DC
Submitted: 2026-11-12
Updated: 2026-11-12
Code: https://github.com/tile-ai/tilert
Terminology
Sources
- FLUX: Fast Software-based Communication Overlap On GPUs Through Kernel Fusion
- cuDNN: Efficient Primitives for Deep Learning
- VDCores: Resource Decoupled Programming and Execution for Asynchronous GPU
- DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models
- Eliminating Hidden Serialization in Multi-Node Megakernel Communication
- Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
- Mixture-of-Kittens: MoE Megakernel for NVL72s
- Triton-distributed: Programming Overlapping Kernels on Distributed AI Systems with the Triton Compiler
- DITRON: Distributed Multi-level Tiling Compiler for Parallel Tensor Programs
Related papers
- BatchGen: An Architecture for Scalable and Efficient Batch Inference
- Seamless Reconfiguration for DAG-BFT (Technical Report)
- ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
- iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
- TIRx: A Unified Tile Primitive ML Compiler Across Heterogeneous Backends
- AgentWeave: Efficient Distributed Agent Serving via Flow Decomposition