LLM Inference Unveiled: Survey and Roofline Model Insights

Fields

id

sources/flop-llm-inference-roofline-survey

URL

https://arxiv.org/abs/2402.16363

Publisher

arXiv (Yuan, Shang, Zhou et al.)

Kind

paper

Published

2024-02-26

Accessed

2026-10-03

notes

Roofline-based framework for LLM inference that explicitly answers why LLMs are memory-bound and how to pick hardware. Open-sourced tool LLM-Viewer for visualizing arithmetic intensity against hardware ceilings.

Notes

Roofline-based framework for LLM inference that explicitly answers why LLMs are memory-bound and how to pick hardware. Open-sourced tool LLM-Viewer for visualizing arithmetic intensity against hardware ceilings.

Cross-references

This record references no other records.