Skip to main content

GPU Optimization

Explore machine learning papers and reviews related to GPU Optimization. Find insights, analysis, and implementation details.

  • Tagged with
  • 9 papers
Back to all papers

Papers Related to GPU Optimization

OSDI 2020

HiveD: Sharing a GPU Cluster for Deep Learning with Guarantees

Hanyu Zhao, Zhenhua Han, +9

HiveD reserves groups of GPUs that sit together, not GPU counts, so a tenant never queues longer in a shared cluster than in a private one of the same size.

arXiv 2019

Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism

Mohammad Shoeybi, Mostofa Patwary, +4

Megatron-LM splits each transformer layer across GPUs with two all-reduces forward and two backward, and trains an 8.3B-parameter GPT-2 on 512 V100s.

arXiv 2025

Collective Communication for 100k+ GPUs

Min Si, Pavan Balaji, +37

NCCLX is Meta's communication stack for Llama 4: host-driven, zero-copy collectives that start a 96K-GPU job 11x faster and cut decode time by up to 83%.

arXiv 2026

Every µs Matters: Achieving Near Speed-of-Light Latency in GPU Collectives

Siyuan Shen, Anton Korzh, +11

How the Every µs Matters paper brings small GPU AllReduce operations close to the hardware speed-of-light bound: it measures a 1.404 µs floor on GB200, removes memory barriers with LL, sentinel and LL128 atomic signalling, and cuts vLLM inter-token latency by up to 13%.

OSDI 2025

NanoFlow: Towards Optimal Large Language Model Serving Throughput

Kan Zhu, Yufei Gao, +14

How NanoFlow raises LLM serving throughput by overlapping compute, memory and network work inside a single GPU: it shows serving is compute-bound, splits each batch into nano-batches, and shares SMs between concurrent kernels, reaching 1.91x the throughput of TensorRT-LLM.

NeurIPS 2022

FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness

Tri Dao, Daniel Y. Fu, +3

How FlashAttention computes exact attention with linear memory by tiling Q, K, and V into SRAM-resident blocks and fusing the softmax, avoiding the quadratic HBM cost of materializing the full attention matrix.

Mastodon