diff options
| author | bors[bot] <26634292+bors[bot]@users.noreply.github.com> | 2020-04-16 13:50:19 +0100 |
|---|---|---|
| committer | GitHub <[email protected]> | 2020-04-16 13:50:19 +0100 |
| commit | 7d60a446fca0923f698c4b8d10236ae246012f4e (patch) | |
| tree | 266feed101e657c3cc494f19843c65e19f29fe31 /crates/ra_hir_ty/src/infer | |
| parent | 364415b7d66bc9d42f21181d7f642e9f911c4711 (diff) | |
| parent | 39fe3a6486a2cbdf00bce8bd4861a900e0ff5811 (diff) | |
Merge #3990
3990: Switch to Chalk recursive solver r=matklad a=flodiebold
Before:
```
Expressions of unknown type: 5526 (3%)
Expressions of partially unknown type: 5415 (3%)
```
After:
```
Expressions of unknown type: 4600 (2%)
Expressions of partially unknown type: 4645 (2%)
```
On the other hand,
```
'./target/release/rust-analyzer analysis-stats -q . # old solver' ran
1.24 ± 0.04 times faster than 'rust-analyzer analysis-stats -q . # new solver'
```
I think part of this just comes from the fact that we're inferring more types now; but apart from that, it should be possible to improve the performance a bunch, and I'll make looking into that a priority.
Co-authored-by: Florian Diebold <[email protected]>
Diffstat (limited to 'crates/ra_hir_ty/src/infer')
| -rw-r--r-- | crates/ra_hir_ty/src/infer/unify.rs | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/crates/ra_hir_ty/src/infer/unify.rs b/crates/ra_hir_ty/src/infer/unify.rs index ac25f8a80..5f6cea8d3 100644 --- a/crates/ra_hir_ty/src/infer/unify.rs +++ b/crates/ra_hir_ty/src/infer/unify.rs | |||
| @@ -32,6 +32,7 @@ where | |||
| 32 | var_stack: Vec<TypeVarId>, | 32 | var_stack: Vec<TypeVarId>, |
| 33 | } | 33 | } |
| 34 | 34 | ||
| 35 | #[derive(Debug)] | ||
| 35 | pub(super) struct Canonicalized<T> { | 36 | pub(super) struct Canonicalized<T> { |
| 36 | pub value: Canonical<T>, | 37 | pub value: Canonical<T>, |
| 37 | free_vars: Vec<InferTy>, | 38 | free_vars: Vec<InferTy>, |