carryless_mul: mention the base
Arithmetic operations do not typically care about the base that is used to represent numbers, but this one does. Mentioning that makes it easier to understand the operation, I think.
Cc @folkertdev
make `rustc_allow_const_fn_unstable` an actual `rustc_attrs` attribute
It is already named like one, but used to have its own feature gate, which this PR now removes in favor of just using `#![feature(rustc_attrs)]`.
Most of the diff is just the line number changes in `malformed-attrs.stderr`.
Add documentation note about signed overflow direction
In https://github.com/rust-lang/rust/issues/151989#issuecomment-3845282666 I noticed that signed overflow direction can be determined by returned wrapped value. It is not very obvious (especially, assuming additional `carry: bool` summand), but it is important if we want to add new leading (signed) limb to big integer in this case.
Examples for small summands `x, y: i8` with result extension:
| x | y | overflow | result as (u8, i8) |
| ---- | ---- | -------- | ------------------ |
| -1 | -128 | true | (127, -1) |
| 0 | -1 | false | (255, -1) |
| 2 | 2 | false | (4, 0) |
| 127 | 1 | true | (128, 0) |
Here is general proof.
1. Set $s=2^{N-1}$ and let's say `iN::carrying_add(x, y, c)` returns `(result, true)` then
$$
\mathrm{result}=\begin{cases}
x + y + c + 2s,& x + y + c \le -s-1,\\
x+y+c-2s,& x+y+c\ge s.
\end{cases}
$$
First case is overflowing below `iN::MIN` and we have
$$
\mathrm{result}\ge -s-s+0+2s =0;\qquad
\mathrm{result}=x + y + c + 2s\le -s-1+2s = s - 1,
$$
so we obtain $[0; s-1]$ which is exactly range of non-negative `iN`.
Second case is overflowing above `iN::MAX` and
$$
\mathrm{result}=x+y+c-2s\ge s-2s =-s;\qquad
\mathrm{result}\le s-1 + s-1+1-2s = -1,
$$
that is, $[-s,-1]$ which is exactly range of negative `iN`.
2. Now suppose that `iN::borrowing_sub(x,y,b)` returns `(result, true)` then
$$
\mathrm{result}=\begin{cases}
x - y - b + 2s,& x - y - b \le -s-1,\\
x - y - b - 2s,& x - y - b\ge s.
\end{cases}
$$
First case is overflowing below `iN::MIN` and we have
$$
\mathrm{result}\ge -s-(s-1)-1+2s =0;\qquad
\mathrm{result}=x - y - b + 2s\le -s-1+2s = s - 1.
$$
Second case is overflowing above `iN::MAX` and
$$
\mathrm{result}=x-y-b-2s\ge s-2s =-s;\qquad
\mathrm{result}\le s-1 - (-s) - 0 - 2s = -1.
$$
Improve `VaList` stdlib docs
tracking issue: https://github.com/rust-lang/rust/issues/44930
Some improvements to the `VaList` documentation, at least adding an example. We should link to the reference for c-variadic functions once stable. I've tried to call out explicitly that the type is meant for sending over the FFI boundary.
r? workingjubilee
cc @tgross35
Implement accepted ACP for `MAX_EXACT_INTEGER` and `MIN_EXACT_INTEGER`
on `f16`, `f32`, `f64`, and `f128`
Add tests to `coretests/tests/floats/mod.rs`
Disable doc tests for i586 since float<->int casts return incorrect
results
support c-variadic functions in `rustc_const_eval`
tracking issue: https://github.com/rust-lang/rust/issues/44930
The new `GlobalAlloc::VaList` is used to create an `AllocId` that represents the variable argument list of a frame. The allocation itself does not store any data, all we need is the unique identifier.
The actual variable argument list is stored in `Memory`, and keyed by the `AllocId`. The `Frame` also stores this `AllocId`, so that when a frame is popped, it can deallocate the variable arguments.
At "runtime" a `VaList` value stores a pointer to the global allocation in its first bytes. The provenance on this pointer can be used to retrieve its `AllocId`, and the offset of the pointer is used to store the index of the next argument to read from the variable argument list.
Miri does not yet support `va_arg`, but I think that can be done separetely?
r? @RalfJung
cc @workingjubilee
Pass alignments through the shim as `Alignment` (not `usize`)
We're using `Layout` on both sides, so might as well skip the transmutes back and forth to `usize`.
The mir-opt test shows that doing so allows simplifying the boxed-slice drop slightly, for example.
Improve write! and writeln! error when called without destination
Fixesrust-lang/rust#152493
Adds catch-all arms to `write!` and `writeln!` macros so that calling them without a destination (e.g., `write!("S")` instead of `write!(f, "S")`) gives a clear error instead of the cryptic "unexpected end of macro invocation" pointing at macro internals.
r? @estebank
implement `carryless_mul`
tracking issue: https://github.com/rust-lang/rust/issues/152080
ACP: https://github.com/rust-lang/libs-team/issues/738
This defers to LLVM's `llvm.clmul` when available, and otherwise falls back to a method from the `polyval` crate ([link](https://github.com/RustCrypto/universal-hashes/blob/master/polyval/src/field_element/soft/soft64.rs)).
Some things are missing, which I think we can defer:
- the ACP has some discussion about additional methods, but I'm not sure exactly what is wanted or how to implement it efficiently
- the SIMD intrinsic is not yet `const` (I think I ran into a bootstrapping issue). That is fine for now, I think in `stdarch` we can't really use this intrinsic at the moment, we'd only want the scalar version to replace some riscv intrinsics.
- the SIMD intrinsic is not implemented for the gcc and cranelift backends. That should be reasonably straightforward once we have a const eval implementation though.
Do not require `'static` for obtaining reflection information.
tracking issue rust-lang/rust#142577
This does not affect the stable `TypeId::of`, as that has its own `'static` bound.
But it will allow obtaining `TypeId`s for non-static types via the reflection API. To obtain such a `TypeId` for any type, just use `Type::of::<(T,)>().kind` to extract the first field of a tuple.
This effectively reintroduces rust-lang/rust#41875, which @rust-lang/lang decided against allowing back in 2018 due to lack of sound use cases. We will thus need to have a T-lang meeting specifically about `TypeId` for non-static types before *stabilizing* any part of reflection (in addition to T-lang meetings about reflection in general). I'm adding an explicit point about this to the tracking issue.
cc @scottmcm @joshtriplett @9SonSteroids @SpriteOvO @izagawd @BD103
UnsafePinned: implement opsem effects of UnsafeUnpin
This implements the next step for https://github.com/rust-lang/rust/issues/125735: actually making `UnsafePinned` have special opsem effects by suppressing the `noalias` *even if* the type is wrapped in an `Unpin` wrapper.
For backwards compatibility we also still keep the `Unpin` hack, i.e. a type must be both `Unpin` and `UnsafeUnpin` to get `noalias`.
Optimize indexing slices and strs with inclusive ranges
Instead of separately checking for `end == usize::MAX` and `end + 1 > slice.len()`, we can check for `end >= slice.len()`. Also consolidate all the str indexing related panic functions into a single function which reports the correct error depending on the arguments, as the slice indexing code already does.
The downside of all this is that the panic message is slightly less specific when trying to index with `[..=usize::MAX]`: instead of saying "attempted to index str up to maximum usize" it just says "end byte index {end} out of bounds". But this is a rare enough case that I think it is acceptable
We're using `Layout` on both sides, so might as well skip the transmutes back and forth to `usize`.
The mir-opt test shows that doing so allows simplifying the boxed-slice drop slightly, for example.
Start using pattern types in libcore (NonZero and friends)
part of rust-lang/rust#136006
This PR only changes the internal representation of `NonZero`, `NonMax`, ... and other integral range types in libcore. This subsequently affects other types made up of it, but nothing really changes except that the field of `NonZero` is now accessible safely in contrast to the `rustc_layout_scalar_range_start` attribute, which has all kinds of obscure rules on how to properly access its field.
Support ADT types in type info reflection
Tracking issue: rust-lang/rust#146922 `#![feature(type_info)]`
This PR supports ADT types for feature `type_info` reflection.
(It's still a draft PR, with implementation in progress)
Note that this PR does not take SemVer into consideration (I left a FIXME comment). As discussed earlier ([comment](https://github.com/rust-lang/rust/pull/146923#discussion_r2372249477)), this requires further discussion. However, I hope we could get an initial implementation to land first, so we can start playing with it.
### Progress / Checklist
- [x] Struct support.
- [x] Enum
- [x] Union
- [x] Generics
- [ ] ~Methods~ Implemented and to be implemented in other PRs
- [ ] ~Traits~ Implemented and to be implemented in other PRs
- [x] Rebasing PR to `main` branch
~~(It's currently based on PR rust-lang/rust#151123, so here's an extra commit)~~
- [x] Cleanup and Rebase.
- [x] Fix field info for references. (see [comment](https://github.com/rust-lang/rust/pull/151142#discussion_r2777920512))
r? @oli-obk
Rename `into_range` to `try_into_slice_range`:
- Prepend `try_` to show that it returns `None` on error, like `try_range`
- add `_slice` to make it consistent with `into_slice_range`
The panic message when slicing a string with a negative length range (eg
`"abcdef"[4..3]`) is confusing: it gives the condition that failed to hold,
whilst all the other panic messages give the condition that did hold.
Before:
begin <= end (4 <= 3) when slicing `abcdef`
After:
begin > end (4 > 3) when slicing `abcdef`
Replace `self.end() == usize::MAX` and `self.end() + 1 > slice.len()`
with `self.end() >= slice.len()`. Same reasoning as previous commit.
Also consolidate the str panicking functions into function.
The checks for `self.end() == usize::MAX` and `self.end() + 1 > slice.len()`
can be replaced with `self.end() >= slice.len()`, since
`self.end() < slice.len()` implies both
`self.end() <= slice.len()` and
`self.end() < usize::MAX`.
Align `ArrayWindows` trait impls with `Windows`
With `slice::ArrayWindows` getting ready to stabilize in 1.94, I noticed that it currently has some differences in trait implementations compared to `slice::Windows`, and I think we should align these.
- Remove `derive(Copy)` -- we generally don't want `Copy` for iterators at all, as this is seen as a footgun (e.g. rust-lang/rust#21809). This is obviously a breaking change though, so we should only remove this if we also backport the removal before it's stable. Otherwise, it should at least be replaced by a manual impl without requiring `T: Copy`.
- Manually `impl Clone`, simply to avoid requiring `T: Clone`.
- `impl FusedIterator`, because it is trivially so. The `since = "1.94.0"` assumes we'll backport this, otherwise we should change that to the "current" placeholder.
- `impl TrustedLen`, because we can trust our implementation.
- `impl TrustedRandomAccess`, because the required `__iterator_get_unchecked` method is straightforward.
r? libs-api
@rustbot label beta-nominated
(at least for the `Copy` removal, but we could be more selective about the rest).
Stop having two different alignment constants
Now that there's a `<T as SizedTypeProperties>::ALIGNMENT` constant, `Alignment::of` can use that instead of an inline constant, like how `Layout::new` uses the constant from `SizedTypeProperties`.
Reword the caveats on `array::map`
Thanks to #107634 and some improvements in LLVM (particularly [`dead_on_unwind`](https://llvm.org/docs/LangRef.html#parameter-attributes)), the method actually optimizes reasonably well now.
So focus the discussion on the fundamental ordering differences where the optimizer might never be able to fix it because of the different behaviour, and keep encouraging `Iterator::map` where an array wasn't actually ever needed.
