Feature Request:

[mahald]

Prerequisites

• I am running the latest code. Mention the version if possible as well.
• I carefully followed the README.md.
• I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
• I reviewed the Discussions, and have a new and useful enhancement to share.

Feature Description

there is a truboquant implementation that has zero performance hit on tubo4: https://github.com/test1111111111111112/llama-cpp-turboquant-gemma4 have u seen that ? hope the faster one gets merged to master for turbo4. i sell testet on a RTX 4080 Laptop gpt and with that version i get 80 t/s while with this version here i get 65 t/s only.

Motivation

having zero performance hit on turbo4

Possible Implementation

No response