These functions are unstable, but because they're inherent they still
introduce conflicts with stable trait functions in crates. Temporarily
rename them to fix these conflicts, until we can resolve those conflicts
in a better way.
Add carrying_add, borrowing_sub, widening_mul, carrying_mul methods to integers
This comes in part from my own attempts to make (crude) big integer implementations, and also due to the stalled discussion in [RFC 2417](https://github.com/rust-lang/rfcs/pull/2417). My understanding is that changes like these are best offered directly as code and then an RFC can be opened if there needs to be more discussion before stabilisation. Since all of these methods are unstable from the start, I figured I might as well offer them now.
I tried looking into intrinsics, messed around with a few different implementations, and ultimately concluded that these are "good enough" implementations for now to at least put up some code and maybe start bikeshedding on a proper API for these.
For the `carrying_add` and `borrowing_sub`, I tried looking into potential architecture-specific code and realised that even using the LLVM intrinsics for `addcarry` and `subborrow` on x86 specifically, I was getting exactly the same assembly as the naive implementation using `overflowing_add` and `overflowing_sub`, although the LLVM IR did differ because of the architecture-specific code. Longer-term I think that they would be best suited to specific intrinsics as that would make optimisations easier (instructions like add-carry tend to use implicit flags, and thus can only be optimised if they're done one-after-another, and thus it would make the most sense to have compact intrinsics that can be merged together easily).
For `widening_mul` and `carrying_mul`, for now at least, I simply cast to the larger type and perform arithmetic that way, since we currently have no intrinsic that would work better for 128-bit integers. In the future, I also think that some form of intrinsic would work best to cover that case, but for now at least, I think that they're "good enough" for now.
The main reasoning for offering these directly to the standard library even though they're relatively niche optimisations is to help ensure that the code generated for them is optimal. Plus, these operations alone aren't enough to create big integer implementations, although they could help simplify the code required to do so and make it a bit more accessible for the average implementor.
That said, I 100% understand if any or all of these methods are not desired simply because of how niche they are. Up to you. 🤷🏻
Make wrapping_neg() use wrapping_sub(), #[inline(always)]
This is a follow-up change to the fix for #75598. It simplifies the implementation of wrapping_neg() for all integer types by just calling 0.wrapping_sub(self) and always inlines it. This leads to much less assembly code being emitted for opt-level≤1 and thus much better performance for debug-compiled code.
Background is [this discussion on the internals forum](https://internals.rust-lang.org/t/why-does-rust-generate-10x-as-much-unoptimized-assembly-as-gcc/14930).
Document math behind MIN/MAX consts on integers
Currently the documentation for `[integer]::{MIN, MAX}` doesn't explain where the constants come from. This documents how the values of those constants are related to powers of 2.
This is a follow-up change to the fix for #75598. It simplifies the implementation of wrapping_neg() for all integer types by just calling 0.wrapping_sub(self) and always inlines it. This leads to much less assembly code being emitted for opt-level≤1.
I tried to add it only to methods which return results of intrinsics and don't have any branching.
Branching could made performance of debug builds (`-Copt-level=0`) worse.
Main goal of changes is allowing wider optimizations in `-Copt-level=1`.
Closes: https://github.com/rust-lang/rust/issues/75598
This commit fixes the statement of the inequality that the Euclidean
remainder satisfies. (The remainder is guaranteed to be less than
abs(rhs), not rhs.) It also rewords the documentation to make it a
little easier to read.
Bump stabilization version for const int methods
These methods missed the beta cutoff. See #80962 for details.
`@rustbot` modify labels to +A-const-fn, +A-intrinsics
r? `@m-ou-se`
Stabilize remaining integer methods as `const fn`
This pull request stabilizes the following methods as `const fn`:
- `i*::checked_div`
- `i*::checked_div_euclid`
- `i*::checked_rem`
- `i*::checked_rem_euclid`
- `i*::div_euclid`
- `i*::overflowing_div`
- `i*::overflowing_div_euclid`
- `i*::overflowing_rem`
- `i*::overflowing_rem_euclid`
- `i*::rem_euclid`
- `i*::wrapping_div`
- `i*::wrapping_div_euclid`
- `i*::wrapping_rem`
- `i*::wrapping_rem_euclid`
- `u*::checked_div`
- `u*::checked_div_euclid`
- `u*::checked_rem`
- `u*::checked_rem_euclid`
- `u*::div_euclid`
- `u*::overflowing_div`
- `u*::overflowing_div_euclid`
- `u*::overflowing_rem`
- `u*::overflowing_rem_euclid`
- `u*::rem_euclid`
- `u*::wrapping_div`
- `u*::wrapping_div_euclid`
- `u*::wrapping_rem`
- `u*::wrapping_rem_euclid`
These can all be implemented on the current stable (1.49). There are two unstable details: const likely/unlikely and unchecked division/remainder. Both of these are for optimizations, and are in no way required to make the methods function; there is no exposure of these details publicly. Per comments below, it seems best practice is to stabilize the intrinsics. As such, `intrinsics::unchecked_div` and `intrinsics::unchecked_rem` have been stabilized as `const` as part of this pull request as well. The methods themselves remain unstable.
I believe part of the reason these were not stabilized previously was the behavior around division by 0 and modulo 0. After testing on nightly, the diagnostic for something like `const _: i8 = 5i8 % 0i8;` is similar to that of `const _: i8 = 5i8.rem_euclid(0i8);` (assuming the appropriate feature flag is enabled). As such, I believe these methods are ready to be stabilized as `const fn`.
This pull request represents the final methods mentioned in #53718. As such, this PR closes#53718.
`@rustbot` modify labels to +A-const-fn, +T-libs
Stabilize `unsigned_abs`
Resolves#74913.
This PR stabilizes the `i*::unsigned_abs()` method, which returns the absolute value of an integer _as its unsigned equivalent_. This has the advantage that it does not overflow on `i*::MIN`.
I have gone ahead and used this in a couple locations throughout the repository.
Integer types have a `count_ones` method that end up calling
`intrinsics::ctpop`.
On some architectures, that intrinsic is translated as a corresponding
CPU instruction know as "popcount" or "popcnt".
This PR makes it so that searching for those names in rustdoc shows those methods.
CC https://blog.rust-lang.org/2020/11/19/Rust-1.48.html#adding-search-aliases
Also stabilize constctlz for const ctlz_nonzero.
The public methods stabilized const by this commit are:
* `{i*,u*}::checked_pow`
* `{i*,u*}::saturating_pow`
* `{i*,u*}::wrapping_pow`
* `{i*,u*}::overflowing_pow`
* `{i*,u*}::pow`
* `u*::next_power_of_two`
* `u*::checked_next_power_of_two`
* `u*::wrapping_next_power_of_two` (the method itself is still unstable)
