The paper investigates how to optimally partition streaming inference models into megakernels and conventional graph components. It introduces a framework that predicts the best megakernel composition without full end‑to‑end runs, showing that multiple megakernels or a fully conventional graph often outperform a single megakernel across various models and batch sizes. The study defines four model categories based on megakernel stability and demonstrates significant latency and throughput gains on an H100 NVL GPU using a three‑megakernel design.
We address a fundamental question: How far should kernel fusion extend and which boundaries are worth preserving? Unlike the general convention of using one monolith megakernel at lower batch sizes or...
arXiv:2606. 28565v1 Announce Type: cross Abstract: As large language models (LLMs) move into production serving, practitioners must rapidly evaluate inference performance across diverse hardware, models, and serving parameters to meet cost and latency targets.
The paper investigates how a single SM utilization metric can misrepresent the true workload of large language model (LLM) inference on Nvidia Hopper GPUs. By profiling vLLM with FlashAttention‑3 and cuBLASLt on an H100 NVL across various phases (cold prefill, warm prefill, and decode) and varying sequence length and batch size, the authors replace the single utilization figure with eight detailed counter‑validated views. These views, tied to specific Nsight Compute counters or formulas, reveal how factors such as fragment fill, occupancy limits, stall signatures, wave quantization, and kernel selection create utilization gaps across four production models and six per‑layer kernel roles.
arXiv:2609.13585v1 Announce Type: cross Abstract: Communication has become a bottleneck in distributed training and inference of large models. Overlapping communication with computation at the granul...
arXiv:2512. 22219v2 Announce Type: replace-cross Abstract: We introduce Mirage Persistent Kernel (MPK), the first compiler and runtime system that automatically transforms multi-GPU model inference into a single high-performance mega-kernel.
arXiv:2607. 23264v1 Announce Type: cross Abstract: Fine-grained, device-initiated communication lets persistent GPU kernels in distributed diffusion transformer (DiT) inference issue remote stores and overlap data movement with Tensor Core computation.
arXiv:2609.13592v1 Announce Type: cross Abstract: GPU memory bandwidth and capacity limit throughput in large language model (LLM) inference. The GPU memory system consists of a primary tier of high-...
arXiv:2608.28044v1 Announce Type: cross Abstract: Large language model (LLM) inference serving is priced by tokens, but GPU energy is consumed over inference windows. This accounting mismatch makes t...
arXiv:2607. 19539v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) architectures increase model capacity without proportionally increasing computation cost and have become a key building block for scaling large language models (LLMs) to trillion-parameter regimes.
BitMTP is an algorithm‑system co‑design framework that combines multi‑token prediction (MTP) with ultra‑low‑bit large language model backbones to accelerate decoding while maintaining quality. It introduces joint next‑token and MTP objectives, multiple decoding modes (confidence‑only, speculative‑only, hybrid), and an accepted‑prefix alignment mechanism that keeps causal key‑value boundaries and CUDA Graph replay states consistent. Experiments on NVIDIA H20, RTX 3090, and Jetson Orin show speedups ranging from 2.0× to 3.3× across various ultra‑low‑bit models without sacrificing task performance.
arXiv:2605. 25645v2 Announce Type: replace-cross Abstract: We present the first end-to-end demonstration of fine-tuning and serving Google's Gemma 4 31B model on TPU hardware, providing an empirical comparison of TPU and GPU platforms for large language model adaptation.
arXiv:2604. 26968v2 Announce Type: replace-cross Abstract: Key-value (KV) cache memory management is the primary bottleneck limiting throughput and cost-efficiency in large-scale GPU inference serving.