Skip to content

Conversation

@IMbackK
Copy link
Collaborator

@IMbackK IMbackK commented Jan 28, 2025

loops with bounds not known at compile time indeed can not be unrolled.

when ncols_template == 0, the bounds of the loop are not constexpr, thus llvm cant unroll the loops here. This causes annoying repeating warnings:

llama.cpp/ggml/src/ggml-cuda/softmax.cu:17:24: warning: loop not unrolled: the optimizer was unable to perform the requested transformation; the transformation might be disabled or specified as part of an unsupported transformation ordering [-Wpass-failed=transform-warning]

so lets suppress the this warning in this position as we know these loops cant be unrolled but we want to keep the pragma as it unrolls the loop just fine in the ncols_template != 0 case.

@github-actions github-actions bot added Nvidia GPU Issues specific to Nvidia GPUs ggml changes relating to the ggml tensor library for machine learning labels Jan 28, 2025
@JohannesGaessler
Copy link
Collaborator

Please add a brief comment explaining this to the code.

@IMbackK
Copy link
Collaborator Author

IMbackK commented Jan 28, 2025

done

…not known at compiletime indeed can not be unrolled.
Copy link
Collaborator

@JohannesGaessler JohannesGaessler left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

CI didn't finish yet but I would be very surprised if a warning suppression were to cause issues.

@IMbackK IMbackK merged commit be5ef79 into ggml-org:master Jan 28, 2025
45 checks passed
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 Nvidia GPU Issues specific to Nvidia GPUs

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants