Create NormalizeTy query
As part of the effort to solve #44891, I've created the normalize_ty query.
As outlined in the issue this meant:
- renamed `normalize_associated_type()` to `normalize_associated_type_in()`
- created the `normalize_ty` query
- substituted the use of memoize with the query
This PR is not ready. While running tests, one of the incremental ones failed. [This](https://pastebin.com/vGhH6bv6) is the error I got.
Move Generics from MethodSig to TraitItem and ImplItem
As part of `rust-impl-period/WG-compiler-traits`, we want to "lift" `Generics` from `MethodSig` into `TraitItem` and `ImplItem`. This is in preparation for adding associated type generics. (https://github.com/rust-lang/rust/issues/44265#issuecomment-331172238)
Currently this change is only made in the AST. In the future, it may also impact the HIR. (Still discussing)
To understand this PR, it's probably best to start from the changes to `ast.rs` and then work your way to the other files to understand the far reaching effects of this change.
r? @nikomatsakis
This commit moves compression of the bytecode from the `link` module to the
`write` module, namely allowing it to be (a) cached by incremental compilation
and (b) produced in parallel. The parallelization may show up as some nice wins
during normal compilation and the caching in incremental mode should be
beneficial for incremental compiles! (no more need to recompress the entire
crate's bitcode on all builds)
incr.comp.: Use 128bit SipHash for fingerprinting
This PR switches incr. comp. result fingerprinting from 128 bit BLAKE2 to 128 bit SipHash. When we started using BLAKE2 for fingerprinting, the 128 bit version of SipHash was still experimental. Now that it isn't anymore we should be able to get a nice performance boost without significantly increasing collision probability.
~~I'm going to start a try-build for this, so we can gauge the performance impact before merging (hence the `WIP` in the title).~~
EDIT: Performance improvements look as expected. Tests seem to be passing.
Fixes#41215.
Mark block exits as reachable if the block can break.
This only happens when desugaring `catch` expressions for now, but regular blocks (in HIR) can be broken from - respect that when doing reachability analysis.
Fixes#45124.
Refactor `ensure` and `try_get_with`
There was a bit of code shared between `try_get_with` and `ensure`, after I
added `ensure`. I refactored that shared code into a query-agnostic method
called `read_node_index`.
The new method `read_node_index` will attempt to find the node
index (`DepNodeIndex`) of a query. When `read_node_index` finds the
`DepNodeIndex`, it marks the current query as a reader of the node it's
requesting the index of.
This is used by `try_get_with` and `ensure` as it elides the unimportant (to
them) details of if the query is invalidated by previous changed computation (Red)
or new and if they had to mark the query green. For both `try_get_with` and
`ensure`, they just need to know if they can lookup the results or have to
reevaluate.
@nikomatsakis this is the [refactor we discussed](https://github.com/rust-lang/rust/pull/45228#discussion_r144577392) in the comment thread of #45228
remove or encapsulate the remaining non-query data in tcx
I wound up removing the existing cache around inhabitedness since it didn't seem to be adding much value. I reworked const rvalue promotion, but not that much (i.e., I did not split the computation into bits, as @eddyb had tossed out as a suggestion). But it's now demand driven, at least.
cc @michaelwoerister -- see the `forbid_reads` change in last commit
r? @eddyb -- since the trickiest of this PR is the work on const rvalue promotion
cc #44137
The cache was broken anyhow and this computation doesn't look that
expensive. These public accessors could potentially become queries,
but we'd have to add some more complex logic around lift. I'd prefer
to have some test cases to profile with before doing that.
Fixes#44402.
rustc: Remove `used_mut_nodes` from `TyCtxt`
This updates the borrowck query to return a result, and this result is then used
to incrementally check for unused mutable nodes given sets of all the used
mutable nodes.
Closes#42384
There was a bit of code shared between `try_get_with` and `ensure`, after I
added `ensure`. I refactored that shared code into a query-agnostic method
called `read_node_index`.
The new method `read_node_index` will attempt to find the node
index (`DepNodeIndex`) of a query. When `read_node_index` finds the
`DepNodeIndex`, it marks the current query as a reader of the node it's
requesting the index of.
This is used by `try_get_with` and `ensure` as it elides the unimportant (to
them) details of if the query is invalidated by previous changed computation (Red)
or new and if they had to mark the query green. For both `try_get_with` and
`ensure`, they just need to know if they can lookup the results or have to
reevaluate.
incr.comp.: Introduce `ensure` and `ensure` typeck_tables_of
Resolves#45210
In this Pull Request we introduce the `ensure` query/function. `ensure` has the
semantics and type of the function `Q1` below:
```rust
fn Q1::ensure(K){
Q(K);
}
```
Further, `ensure` avoids the need to load the result from disk (or execute the
provider, if we are not storing the results of Q to disk).
@nikomatsakis
This updates the borrowck query to return a result, and this result is then used
to incrementally check for unused mutable nodes given sets of all the used
mutable nodes.
Closes#42384
Queryify Vtable methods
This query might come with a downside: It converts an iterator to a Vec, which may increase the working set of rustc on programs that use many many traits (I think that's where this is used).
