---
cited_by:
  - '[[thoughts/inference/replayssm-serving#5-pr-reading-order|replayssm §5]]'
date: '2026-11-11'
kind: pr
local:
  - file:///Users/aarnphm/workspace/vllm/vllm/v1/attention/backends/flashinfer.py
merged: '2026-09-11'
number: 50439
source:
  - https://github.com/vllm-project/vllm/blob/4b839c3378c6a40fa613328a9b97069ed83b181b/vllm/v1/attention/backends/flashinfer.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#50439 [Attention] Extend XQA decode support on SM90'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/50439
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/50439
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/50439.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Follow up on [#49718](https://github.com/aarnphm/aarnphm.github.io/issues/49718), which added the dedicated FlashInfer XQA path for SM12x. · full dump: [[thoughts/inference/vllm-prs/50439|#50439]]

