Optimize scalar and scalar pair representations loaded from ByRef in llvm
in https://github.com/rust-lang/rust/pull/105653 I noticed that we were generating suboptimal LLVM IR if we had a `ConstValue::ByRef` that could be represented by a `ScalarPair`. Before https://github.com/rust-lang/rust/pull/105653 this is probably rare, but after it, every slice will go down this suboptimal code path that requires LLVM to untangle a bunch of indirections and translate static allocations that are only used once to read a scalar pair from.
Make `TrustedStep` require `Copy`
All the implementations of the trait already are `Copy`, and this seems to be enough to simplify the implementations enough to make the MIR inliner willing to inline basics like `Range::next`.
r? `@thomcc`
Rollup of 6 pull requests
Successful merges:
- #111558 (Move tests)
- #111827 (Add build instructions for cranelift backend as part of Rust repo)
- #111988 (Make `TyKind: Debug` have less verbose output)
- #112022 (Check nested obligations during coercion unify in new solver)
- #112057 (Suggest correct `self_ty`)
- #112063 (Add a test for issue 110457/incremental ICE with closures with the same span)
Failed merges:
- #112068 (Move tests from `ui/discrim` dir)
r? `@ghost`
`@rustbot` modify labels: rollup
All the implementations of the trait already are `Copy`, and this seems to be enough to simplify the implementations enough to make the MIR inliner willing to inline basics like `Range::next`.
Add a test for issue 110457/incremental ICE with closures with the same span
Closes#110457
It's probably possible to minimize the test case more, considering that we now know the underlying reason for the ICE, but I didn't.
r? `@cjgillot`
Make `TyKind: Debug` have less verbose output
Current `TyKind: Debug` impl is basically unusable for debugging, its too verbose even for verbose debugging 🤣 This PR replaces the debug logic for `TyKind` with a more manual debug impl instead of a hand expanded derived impl. This should help make #107084 more reasonable to land since the output of `Ty: Debug` will be better.
This isn't a fully completed change to the `Debug` impl of `TyKind` as there's still logic from the derive macro for some variants. Some of the variants are also not consisten with the `-Zverbose` printing of `Ty`, ideally `-Zverbose` printing of `Ty` would also just defer to the debug impl instead of having lots of checks in pretty printing. I plan on fixing this in follow up PRs since it seems tricky to do in this one and its already a large PR 😅
Add build instructions for cranelift backend as part of Rust repo
All other instructions assume that user works with separate repository than Rust compiler repository. When one follows default instructions, cranelift codegen tries to use different sys-root and compiler internal crates which leads to compiler errors when building it.
I needed to do all this steps while adding new intrinsic to rustc.
r? bjorn3
Load only the crate header for `locator::crate_matches`
Previously, we used the following info to determine whether to load the crate:
1. The METADATA_HEADER, which includes a METADATA_VERSION constant
2. The embedded rustc version
3. Various metadata in the `CrateRoot`, including the SVH
This worked ok most of the time. Unfortunately, when building locally the rustc version is always
the same because `omit-git-hash` is on by default. That meant that we depended only on 1 and 3, and
we are not very good about bumping METADATA_VERSION (it's currently at 7) so in practice we were
only depending on 3. `CrateRoot` is a very large struct and changes somewhat regularly, so this led
to a steady stream of crashes from trying to load it.
Change the logic to add an intermediate step between 2 and 3: introduce a new `CrateHeader` struct
that contains only the minimum info needed to decide whether the crate should be loaded or not. That
avoids having to load all of `CrateRoot`, which in practice means we should crash much less often.
Note that this works because the SVH should be different between any two dependencies, even if the
compiler has changed, because we use `-Zbinary-dep-depinfo` in bootstrap. See
https://github.com/rust-lang/rust/pull/111329#issuecomment-1538303474 for more details about how the
original crash happened.
Use `Cow` in `{D,Subd}iagnosticMessage`.
Each of `{D,Subd}iagnosticMessage::{Str,Eager}` has a comment:
```
// FIXME(davidtwco): can a `Cow<'static, str>` be used here?
```
This commit answers that question in the affirmative. It's not the most compelling change ever, but it might be worth merging.
This requires changing the `impl<'a> From<&'a str>` impls to `impl From<&'static str>`, which involves a bunch of knock-on changes that require/result in call sites being a little more precise about exactly what kind of string they use to create errors, and not just `&str`. This will result in fewer unnecessary allocations, though this will not have any notable perf effects given that these are error paths.
Note that I was lazy within Clippy, using `to_string` in a few places to preserve the existing string imprecision. I could have used `impl Into<{D,Subd}iagnosticMessage>` in various places as is done in the compiler, but that would have required changes to *many* call sites (mostly changing `&format("...")` to `format!("...")`) which didn't seem worthwhile.
r? `@WaffleLapkin`
Rollup of 5 pull requests
Successful merges:
- #112029 (Recover upon mistyped error on typo'd `const` in const param def)
- #112037 (Add details about `unsafe_op_in_unsafe_fn` to E0133)
- #112039 (compiler: update solaris/illumos to enable tsan support.)
- #112042 (Migrate GUI colors test to original CSS color format)
- #112045 (Followup to #111973)
r? `@ghost`
`@rustbot` modify labels: rollup
Recover upon mistyped error on typo'd `const` in const param def
And add machine-applicable fix for the typo'd `const` keyword.
### Before
```
error: expected one of `,`, `:`, `=`, or `>`, found `N`
--> src/lib.rs:1:18
|
1 | pub fn bar<Const N: u8>() {}
| ^ expected one of `,`, `:`, `=`, or `>`
```
### After This PR
```
error: `const` keyword was mistyped as `Const`
--> test.rs:1:8
|
1 | fn bar<Const N: u8>() {}
| ^^^^^
|
help: use the `const` keyword
|
1 | fn bar<const N: u8>() {}
| ~~~~~
```
Fixes#111941.
Inline derived `hash`
Because most of the other derived functions are inlined: `clone`, `default`, `eq`, `partial_cmp`, `cmp`. The exception is `fmt`, but it tends to not be on hot paths as much.
r? `@ghost`
Each of `{D,Subd}iagnosticMessage::{Str,Eager}` has a comment:
```
// FIXME(davidtwco): can a `Cow<'static, str>` be used here?
```
This commit answers that question in the affirmative. It's not the most
compelling change ever, but it might be worth merging.
This requires changing the `impl<'a> From<&'a str>` impls to `impl
From<&'static str>`, which involves a bunch of knock-on changes that
require/result in call sites being a little more precise about exactly
what kind of string they use to create errors, and not just `&str`. This
will result in fewer unnecessary allocations, though this will not have
any notable perf effects given that these are error paths.
Note that I was lazy within Clippy, using `to_string` in a few places to
preserve the existing string imprecision. I could have used `impl
Into<{D,Subd}iagnosticMessage>` in various places as is done in the
compiler, but that would have required changes to *many* call sites
(mostly changing `&format("...")` to `format!("...")`) which didn't seem
worthwhile.
MIR: opt-in normalization of `BasicBlock` and `Local` numbering
This doesn't matter at all for actual codegen, but after spending some time reading pre-codegen MIR, I was wishing I didn't have to jump around so much in reading post-inlining code.
So this add two passes that are off by default for every mir level, but can be enabled (`-Zmir-enable-passes=+ReorderBasicBlocks,+ReorderLocals`) for humans.
Make `EarlyBinder`'s inner value private
Currently, `EarlyBinder(T)`'s inner value is public, which allows implicitly skipping the binder by indexing into the tuple struct (i.e., `x.0`). `@lcnr` suggested making `EarlyBinder`'s inner value private so users are required to explicitly call `skip_binder` (https://github.com/rust-lang/rust/issues/105779#issuecomment-1549933424) .
This PR makes the inner value private, adds `EarlyBinder::new` for constructing a new instance, and replaces uses of `x.0` with `x.skip_binder()` (or similar). It also adds some documentation to `EarlyBinder::skip_binder` explaining how to skip the binder of `&EarlyBinder<T>` to get `&T` now that the inner value is private (since previously we could just do `&x.0`).
r? `@lcnr`
Use only one shard with a single thread
This changes `Sharded` to only access a single shard using a mask set to `0` when a single thread is used, which leads to cache utilization improvements.
Performance improvement with 1 thread and `cfg(parallel_compiler)`:
<table><tr><td rowspan="2">Benchmark</td><td colspan="1"><b>Before</b></th><td colspan="2"><b>After</b></th></tr><tr><td align="right">Time</td><td align="right">Time</td><td align="right">%</th></tr><tr><td>🟣 <b>clap</b>:check</td><td align="right">1.7402s</td><td align="right">1.7004s</td><td align="right">💚 -2.29%</td></tr><tr><td>🟣 <b>hyper</b>:check</td><td align="right">0.2633s</td><td align="right">0.2550s</td><td align="right">💚 -3.12%</td></tr><tr><td>🟣 <b>regex</b>:check</td><td align="right">0.9716s</td><td align="right">0.9482s</td><td align="right">💚 -2.41%</td></tr><tr><td>🟣 <b>syn</b>:check</td><td align="right">1.5679s</td><td align="right">1.5358s</td><td align="right">💚 -2.05%</td></tr><tr><td>🟣 <b>syntex_syntax</b>:check</td><td align="right">6.0569s</td><td align="right">5.9272s</td><td align="right">💚 -2.14%</td></tr><tr><td>Total</td><td align="right">10.5999s</td><td align="right">10.3666s</td><td align="right">💚 -2.20%</td></tr><tr><td>Summary</td><td align="right">1.0000s</td><td align="right">0.9760s</td><td align="right">💚 -2.40%</td></tr></table>
cc `@SparrowLii`
