Add AMX target-features and `x86_amx_intrinsics` feature flag
This is an effort towards #126622. This adds support for all 5 target-features for `AMX`, and introduces the feature flag `x86_amx_intrinsics`, which would gate these target-features and the yet-to-be-implemented amx intrinsics in stdarch.
Generalize `fn allocator` for Rc/Arc.
Split out from #119761
- For `Rc`/`Arc`, the existing associated `fn`s are changed to allow unsized pointees.
- For `Weak`s, new methods are added.
`````@rustbot````` label +A-allocators
Use rustc-stable-hash in the compiler
Following https://github.com/rust-lang/compiler-team/issues/755 and the release of the crate on crates.io, let's now use it in the compiler and remove the old implementation.
cc `@michaelwoerister`
r? ghost
A `rustc_const_stable` attribute by itself has nonintuitive purpose when
placed in a public module.
Separately, it would probably be okay to rename `const_strlen` to just
`strlen` to make it more clear this is our general-purpose
implementation of strlen now, not something specifically for const
(avoiding confusion like in PR 127444).
Rollup of 6 pull requests
Successful merges:
- #127164 (match lowering: Clarify the main loop of the algorithm)
- #127422 (as_simd: fix doc comment to be in line with align_to)
- #127596 (More suggestion for converting `Option<&Vec<T>>` to `Option<&[T]>`)
- #127607 (compiletest: Better error message for bad `normalize-*` headers)
- #127622 (Mark `builtin_syntax` as internal)
- #127625 (Revert accidental comment deletion)
r? `@ghost`
`@rustbot` modify labels: rollup
This new special case is simpler than the old special case because it
only is used when `dist == 1`. But that's still enough to cover ~98% of
cases. This results in equivalent performance to the old special case,
and identical behaviour as the general case.
The general case at the bottom of `look_ahead` is slow, because it
clones the token cursor. Above it there is a special case for
performance that is hit most of the time and avoids the cloning.
Unfortunately, its behaviour differs from the general case in two ways.
- When within a pair of delimiters, if you look any distance past the
closing delimiter you get the closing delimiter instead of what comes
after the closing delimiter.
- It uses `tree_cursor.look_ahead(dist - 1)` which totally confuses
tokens with token trees. This means that only the first token in a
token tree will be seen. E.g. in a sequence like `{ a }` the `a` and
`}` will be skipped over. Bad!
It's likely that these differences weren't noticed before now because
the use of `look_ahead` in the parser is limited to small distances and
relatively few contexts.
Removing the special case causes slowdowns up of to 2% on a range of
benchmarks. The next commit will add a new, correct special case to
regain that lost performance.
compiletest: Better error message for bad `normalize-*` headers
Follow-up to #126777.
Example of the new error message in context:
```text
---- [ui] tests/ui/rfcs/rfc-2632-const-trait-impl/effects/minicore.rs stdout ----
thread '[ui] tests/ui/rfcs/rfc-2632-const-trait-impl/effects/minicore.rs' panicked at src/tools/compiletest/src/header.rs:1001:13:
couldn't parse custom normalization rule: `normalize-stderr-test ".*note: .*\n\n" -> ""`
help: expected syntax is: `normalize-stderr-test: "REGEX" -> "REPLACEMENT"`
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
```
as_simd: fix doc comment to be in line with align_to
In #121201, the guarantees about `align_offset` and `align_to` were changed. This PR aims to correct the doc comment of `as_simd` to be in line with the new `align_to`.
Tagging #86656 for good measure.
match lowering: Clarify the main loop of the algorithm
Now that we expand or-patterns in a single place in the algorithm, we can move it (back) to the main part of the loop. This makes the call-graph of the main loop rather simple: `match_candidates` has three branches that each call back to `match_candidates`. The remaining tricky part is `finalize_or_candidate`.
I also factored out the whole "process a prefix of the candidates then process the rest" thing which I think helps legibility.
The first two commits are a fix for an indexing mistake I introduced in https://github.com/rust-lang/rust/pull/126553, already sumitted in https://github.com/rust-lang/rust/pull/127028 but feel free to merge this first.
r? `@matthewjasper`
Rollup of 8 pull requests
Successful merges:
- #124599 (Suggest borrowing on fn argument that is `impl AsRef`)
- #127572 (Don't mark `DEBUG_EVENT` struct as `repr(packed)`)
- #127588 (core: Limit remaining f16 doctests to x86_64 linux)
- #127591 (Make sure that labels are defined after the primary span in diagnostics)
- #127598 (Allows `#[diagnostic::do_not_recommend]` to supress trait impls in suggestions as well)
- #127599 (Rename `lazy_cell_consume` to `lazy_cell_into_inner`)
- #127601 (check is_ident before parse_ident)
- #127605 (Remove extern "wasm" ABI)
r? `@ghost`
`@rustbot` modify labels: rollup
Remove extern "wasm" ABI
Remove the unstable `extern "wasm"` ABI (`wasm_abi` feature tracked in #83788).
As discussed in https://github.com/rust-lang/rust/pull/127513#issuecomment-2220410679 and following, this ABI is a failed experiment that did not end up being used for anything. Keeping support for this ABI in LLVM 19 would require us to switch wasm targets to the `experimental-mv` ABI, which we do not want to do.
It should be noted that `Abi::Wasm` was internally used for two things: The `-Z wasm-c-abi=legacy` ABI that is still used by default on some wasm targets, and the `extern "wasm"` ABI. Despite both being `Abi::Wasm` internally, they were not the same. An explicit `extern "wasm"` additionally enabled the `+multivalue` feature.
I've opted to remove `Abi::Wasm` in this patch entirely, instead of keeping it as an ABI with only internal usage. Both `-Z wasm-c-abi` variants are now treated as part of the normal C ABI, just with different different treatment in
adjust_for_foreign_abi.
Allows `#[diagnostic::do_not_recommend]` to supress trait impls in suggestions as well
This commit changes the error reporting mechanism for not implemented traits to skip impl marked as `#[diagnostic::do_not_recommend]` in the help part of the error message ("the following other types implement trait `Foo`:"). The main use case here is to allow crate authors to skip non-meaningful confusing suggestions. A common example for this are fully generic impls on tuples.
Related to https://github.com/rust-lang/rust/issues/51992
r? `@compiler-errors`
Make sure that labels are defined after the primary span in diagnostics
Putting a `#[label]` before a `#[primary_span]` results in that label being overwritten, due to the semantics of `Diagnostic::span` and the fact that labels are stored in the `MultiSpan` of the diagnostic.
This isn't possible to fix in general, since a lot of code actually *relies* in this overwriting behavior (e.g. `rustc_on_unimplemented`). However, it's useful to enforce this for derive-diagnostics, since this is certainly never what you intend to do in a derived diagnostic, where all the fields are meaningful parts of the diagnostic being rendered.
This only matters for `#[label]`, since those are the ones stored in the `MultiSpan` of the error.
We could also make this "just work" by sorting the attrs or processing the primary span attr first, however I think it's kinda pointless to do.
There was 1 case where this mattered, but we literally didn't have a test exercising that diagnostic 🙃
core: Limit remaining f16 doctests to x86_64 linux
On s390x, every use of the f16 data type will currently ICE due to https://github.com/llvm/llvm-project/issues/50374, causing doctest failures on the platform.
Most doctests were already restricted to certain platforms, so fix this by likewise restricting the remaining five.
Suggest borrowing on fn argument that is `impl AsRef`
When encountering a move conflict, on an expression that is `!Copy` passed as an argument to an `fn` that is `impl AsRef`, suggest borrowing the expression.
```
error[E0382]: use of moved value: `bar`
--> f204.rs:14:15
|
12 | let bar = Bar;
| --- move occurs because `bar` has type `Bar`, which does not implement the `Copy` trait
13 | foo(bar);
| --- value moved here
14 | let baa = bar;
| ^^^ value used here after move
|
help: borrow the value to avoid moving it
|
13 | foo(&bar);
| +
```
Fix#41708
binutils 2.40 is required by LLVM 19, as older versions do not
know about the zmmull extension.
I've had to backport some patches to glibc and gcc as well,
as they don't build with binutils 2.40. Alternatively, we could
also switch to glibc 2.35 and gcc 12 (I think). I figured we'd
want to avoid the glibc version change, but if that's fine for
riscv I can go with that instead.
Add `f16` and `f128` as simd types in LLVM
`@sayantn` is working on adding SIMD for `f16` and hitting the `FloatingPointVector` error. This should fix it and unblock adding support for `simd_fma` and `simd_fabs` in stdarch.
