Skip to content

vulkan: optimize Krea2 kernels on RDNA3 - #27480

Closed
pwilkin wants to merge 1 commit into
ggml-org:masterfrom
pwilkin:optimize-krea-vulkan
Closed

vulkan: optimize Krea2 kernels on RDNA3#27480
pwilkin wants to merge 1 commit into
ggml-org:masterfrom
pwilkin:optimize-krea-vulkan

Conversation

@pwilkin

@pwilkin pwilkin commented Aug 21, 2026

Copy link
Copy Markdown
Member

Overview

A series of optimizations for stablediffusion.cpp generation that takes Krea2 8-step generation down from 77 to 29 seconds on Vulkan.

Additional information

Should be stacked on top of #26914

Requirements

Assisted-by: OpenAI Codex
@pwilkin
pwilkin requested review from a team and ggerganov as code owners August 21, 2026 08:58
@0cc4m

0cc4m commented Aug 21, 2026

Copy link
Copy Markdown
Contributor

This needs to be split into multiple PRs, and the slop comments are not acceptable.

@github-actions github-actions Bot added testing Everything test related Vulkan Issues specific to the Vulkan backend ggml changes relating to the ggml tensor library for machine learning labels Aug 21, 2026
@pwilkin

pwilkin commented Aug 21, 2026

Copy link
Copy Markdown
Member Author

So cruel :(

Aight, will 3 parts (MAT_MUL, FATTN, fusion) be OK?

@0cc4m

0cc4m commented Aug 21, 2026

Copy link
Copy Markdown
Contributor

Yes

@pwilkin pwilkin closed this Aug 21, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ggml changes relating to the ggml tensor library for machine learning testing Everything test related Vulkan Issues specific to the Vulkan backend

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants