Open
Conversation
When the Func passed to in()/clone_in() does not directly call the wrapped Func, walk down the call graph and wrap the first direct caller found along each branch. This makes it possible to schedule wrappers across anonymous intermediates (e.g. helper Funcs returned from free-standing C++ functions like the downsamplers in local_laplacian), without needing a handle on every Func in between. Funcs that already directly call the target pass through unchanged, and Funcs with no static path to the target are left alone so existing sequences of in()/clone_in() that rely on later substitution still work. Includes a new correctness test (transitive_in) and uses the new feature in apps/local_laplacian to clone gPyramid[0] into gPyramid[1] (~10% speedup). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
alexreinking
approved these changes
May 1, 2026
Member
alexreinking
left a comment
alexreinking
left a comment
There was a problem hiding this comment.
This is a good feature! Lets you treat intermediate/inlined Funcs a bit more like Exprs. I'd suggest adding an example to the Func Doxygen comments, though.
Member
Author
|
I should add there's a design decision here that I went back and forth on: For the Funcs between *this and the args, do you clone them too? I.e. are we cloning the whole subgraph, or are we just walking back to the Func and cloning that one? I ended up not cloning the whole subgraph because I worried about cases where some of the intermediates are named. If you implicitly clone one of the intermediates without realizing it, and then schedule the original, it will only schedule the other uses of it. Seems like a foot-gun. |
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
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.
When the Func passed to in()/clone_in() does not directly call the wrapped Func, walk down the call graph and wrap the first direct caller found along each branch. This makes it possible to schedule wrappers across anonymous intermediates (e.g. helper Funcs returned from free-standing C++ functions like the downsamplers in local_laplacian), without needing a handle on every Func in between.
Funcs that already directly call the target pass through unchanged, and Funcs with no static path to the target are left alone so existing sequences of in()/clone_in() that rely on later substitution still work.
Includes a new correctness test (transitive_in) and uses the new feature in apps/local_laplacian to clone gPyramid[0] into gPyramid[1] (~10% speedup).
This has been a long-standing annoyance that I'm finally fixing. Feature implemented by claude. I just did the change to local_laplacian to use it.