opencl: add fastdiv and use it in set_rows, ported from cuda #17090
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
github-actions
bot
added
ggml
max-krasnyansky
approved these changes
Nov 10, 2025
Collaborator
max-krasnyansky
left a comment
max-krasnyansky
left a comment
There was a problem hiding this comment.
Tested on Gen5.
I see a little perf bump with llama-3.2-1B-Q4_0
Before:
llama_perf_context_print: prompt eval time = 247.46 ms / 205 tokens ( 1.21 ms per token, 828.41 tokens per second)
llama_perf_context_print: eval time = 1597.53 ms / 63 runs (25.36 ms per token, 39.44 tokens per second)
After:
llama_perf_context_print: prompt eval time = 244.81 ms / 205 tokens ( 1.19 ms per token, 837.39 tokens per second)
llama_perf_context_print: eval time = 1558.21 ms / 63 runs (24.73 ms per token, 40.43 tokens per second)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds fast div/mod that has been in CUDA and Vulkan backends for some time. Hexagon backend is about to have it. This implementation is ported from CUDA. Currently perf stays about the same. I think it is beneficial to have it and we can apply it to more ops in the future.