Refactor/cubecl fusion #2815
Merged
Refactor/cubecl fusion #2815
Conversation
This file contains 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
laggui
approved these changes
Feb 17, 2025
There was a problem hiding this comment.
Any reason why this is alone in a new file? Setting the table for incoming stuff? 👀
We can probably remove this file otherwise
Comment on lines
156
to
182
| // #[derive(CubeLaunch, Default)] | ||
| // /// Global arguments that are used for fusing [element wise operations](ElemwiseOp). | ||
| // pub struct GlobalArgs { | ||
| // pub t_f32: Sequence<Tensor<Line<f32>>>, | ||
| // pub t_f16: Sequence<Tensor<Line<f16>>>, | ||
| // pub t_bf16: Sequence<Tensor<Line<bf16>>>, | ||
| // pub t_i64: Sequence<Tensor<Line<i64>>>, | ||
| // pub t_i32: Sequence<Tensor<Line<i32>>>, | ||
| // pub t_i16: Sequence<Tensor<Line<i16>>>, | ||
| // pub t_i8: Sequence<Tensor<Line<i8>>>, | ||
| // pub t_u64: Sequence<Tensor<Line<u64>>>, | ||
| // pub t_u32: Sequence<Tensor<Line<u32>>>, | ||
| // pub t_u16: Sequence<Tensor<Line<u16>>>, | ||
| // pub t_u8: Sequence<Tensor<Line<u8>>>, | ||
| // pub s_f32: Sequence<f32>, | ||
| // pub s_f16: Sequence<f16>, | ||
| // pub s_bf16: Sequence<bf16>, | ||
| // pub s_i64: Sequence<i64>, | ||
| // pub s_i32: Sequence<i32>, | ||
| // pub s_i16: Sequence<i16>, | ||
| // pub s_i8: Sequence<i8>, | ||
| // pub s_u64: Sequence<u64>, | ||
| // pub s_u32: Sequence<u32>, | ||
| // pub s_u16: Sequence<u16>, | ||
| // pub s_u8: Sequence<u8>, | ||
| // } | ||
|
|
Comment on lines
201
to
228
| // impl<R: Runtime> Default for GlobalArgsLaunch<'_, R> { | ||
| // fn default() -> Self { | ||
| // Self::new( | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // Default::default(), | ||
| // ) | ||
| // } | ||
| // } |
Comment on lines
+186
to
+187
| pub tensors: Sequence<GlobalTensor>, | ||
| pub scalars: Sequence<GlobalScalar>, |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Improve Burn's compilation time significantly :)
burncrate, so we don't need to recompileburn-corewhen working on a backend (better caching).burn-cubecl-fusion, that exports all optimizations used byburn-cubeclwhen fusion is activated.Overall, the fusion part of
burn-cubeclwent from 41s compilation time to 6.5s!burn-cubeclnow takes around 10s and compiles in parallel withburn-cubecl-fusion.