Fold by value, not by reference #660
Merged
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
This abstraction requires folding by-reference instead of by-value. Additionally, it hides a call to `skip_binders` without cleanly encapsulating it, so it probably shouldn't exist at all. I assume it is needed within `rustc`, but this workspace can just use `Binders` directly.
`B` never gets folded, only `P`, and the `B::Result: Debug` is not used
This includes `&T` and `Arc<T>`.
ecstatic-morse
force-pushed
the
fold-by-value
branch
from
4b62b7c to
6944f22
Compare
jackh726
approved these changes
Dec 5, 2020
There was a problem hiding this comment.
♥ Thanks for this. This looks good to me generally, and I definitely am not going to particular about clones. Many of these should be "free" when I::InternedTy: Copy.
(I'm much more inclined to just land this as-is and fix things in-tree than to try to nit-pick as a PR)
|
@bors r+ |
|
📌 Commit 6944f22 has been approved by |
|
☀️ Test successful - checks-actions |
bors
added a commit
that referenced
this pull request
Dec 15, 2020
Optimize `Fold` impls for types on the heap Resolves a FIXME in #660. This PR implements the same optimization used by [`rustc`](https://github.com/rust-lang/rust/blob/5be3f9f10e9fd59ea03816840a6051413fbdefae/compiler/rustc_data_structures/src/functor.rs#L13) to avoid unnecessary heap allocations in `Fold` impls for types like `Box` and `Vec`. Chalk's `Fold` trait is more customizable than the one in `rustc`: it allows one type to be `folded` into a different one as well as for early return via a `Result`. Accordingly, types are checked for layout compatibility (size and alignment) before the optimization is enabled. Types with different layouts (as well as zero-sized types) use the original `Fold` impl. Also, care is taken to avoid leaking resources when failure occurs while folding a `Vec`, since this may happen not only due to panicking but also due to an early return. AFAICT, Chalk doesn't have benchmarks, but the new code executes about 350,000 times while running the test suite, so I suspect there will be some benefit for e.g. `rust-analyzer`. That said, I'm not opposed to waiting for some benchmarking before merging this, since it adds some non-trivial (albeit nicely encapsulated) unsafe code.
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.
Resolves #642.
This might need some more work. I suspect there are unnecessary clones. Without familiarity with the code base I found it hard to choose between changing a function signature versus cloning the argument in some places. It should be pretty close though.