Skip to content

Conversation

@netrunnereve
Copy link
Collaborator

This disables #16977 for old GPUs with no float_controls_rte_fp16 support as that causes the fused pipelines to not get generated. I didn't look into fixing this properly but for now this'll prevent the app from segfaulting when it hits that fusion op.

@netrunnereve netrunnereve requested a review from 0cc4m as a code owner November 10, 2025 02:03
@github-actions github-actions bot added Vulkan Issues specific to the Vulkan backend ggml changes relating to the ggml tensor library for machine learning labels Nov 10, 2025
@netrunnereve netrunnereve changed the title disable rms_norm + mul + rope for old gpus vulkan: disable rms_norm + mul + rope for old gpus Nov 10, 2025
@jeffbolznv
Copy link
Collaborator

LGTM. I was trying to avoid generating additional variants, but obviously forgot to check for that in both places.

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 Vulkan Issues specific to the Vulkan backend

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants