---
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/model_executor/kernels/linear/__init__.py
merged: '2026-08-07'
number: 48735
source:
  - https://github.com/vllm-project/vllm/blob/a801e71cb86ecd45cd8cb8f7d1bc08390f2bd68f/vllm/model_executor/kernels/linear/__init__.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#48735 [Perf] Improve `--linear-backend` filtering'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/48735
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/48735
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/48735.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
The `--linear-backend <name>` currently raises at startup for any linear layer type the backend doesn’t cover, so single-scheme backends are unusable. `flashinfer_b12x` is doubly blocked: its only kernel is absent from the NVFP4… · full dump: [[thoughts/inference/vllm-prs/48735|#48735]]

