Make abs, wrapping_abs, overflowing_abs const functions
#63786
Conversation
|
r? @cramertj (rust_highfive has picked a reviewer for you, use r? to override) |
rust-highfive
added
the
S-waiting-on-review
Centril
added
relnotes
Centril
added
the
needs-fcp
|
@rfcbot fcp merge |
|
Team member @alexcrichton has proposed to merge this. The next step is review by the rest of the tagged team members: No concerns currently listed. Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up! See this document for info about what commands tagged team members can give me. |
rfcbot
added
proposed-final-comment-period
Centril
removed
the
needs-fcp
|
Ping @withoutboats @Kimundi @sfackler in case y'all missed this. :) |
|
🔔 This is now entering its final comment period, as per the review above. 🔔 |
rfcbot
added
final-comment-period
|
The final comment period, with a disposition to merge, as per the review above, is now complete. As the automated representative of the governance process, I would like to thank the author for their work and everyone else who contributed. The RFC will be merged soon. |
rfcbot
added
finished-final-comment-period
|
@bors: r+ |
|
📌 Commit adee559 has been approved by |
bors
added
S-waiting-on-bors
Make `abs`, `wrapping_abs`, `overflowing_abs` const functions This makes `abs`, `wrapping_abs` and `overflowing_abs` const functions like rust-lang#58044 makes `wrapping_neg` and `overflowing_neg` const functions. `abs` is made const by returning `(self ^ -1) - -1` = `!self + 1` = `-self` for negative numbers and `(self ^ 0) - 0` = `self` for non-negative numbers. The subexpression `self >> ($BITS - 1)` evaluates to `-1` for negative numbers and `0` otherwise. The subtraction overflows when `self` is `min_value()`, as we would be subtracting `max_value() - -1`; this is when `abs` should overflow. `wrapping_abs` and `overflowing_abs` make use of `wrapping_sub` and `overflowing_sub` instead of the subtraction operator.
Rollup of 8 pull requests Successful merges: - #63786 (Make `abs`, `wrapping_abs`, `overflowing_abs` const functions) - #63989 (Add Yaah to clippy toolstain notification list) - #64256 (test/c-variadic: Fix patterns on powerpc64) - #64292 (lowering: extend temporary lifetimes around await) - #64311 (lldb: avoid mixing "Hit breakpoint" message with other output.) - #64330 (Clarify E0507 to note Fn/FnMut relationship to borrowing) - #64331 (Changed instant is earlier to instant is later) - #64344 (rustc_mir: buffer -Zdump-mir output instead of pestering the kernel constantly.) Failed merges: r? @ghost
| self | ||
| } | ||
| pub const fn wrapping_abs(self) -> Self { | ||
| (self ^ (self >> ($BITS - 1))).wrapping_sub(self >> ($BITS - 1)) |
There was a problem hiding this comment.
We've had variables for a while, please use them.
There was a problem hiding this comment.
Has it been verified that this optimizes to identical LLVM IR?
There was a problem hiding this comment.
@eddyb Since this has already passed final comment period and been merged, shall I open a new PR to use variables?
There was a problem hiding this comment.
@nikic I can't read LLVM IR; what I did do before submitting the PR was to use llvm-mca, and although the assembly is not identical, the performance measure from llvm-mca was the same, or if I remember well, in some tests the reverse throughput improved by something tiny like 0.1, but not regressed. I only used llvm-mca with i686 or x86-64 instructions, and I don't think I tested all types comprehensively.
There was a problem hiding this comment.
@nikic Hmm, using variables as @eddyb suggested seems to hit two birds with one stone. If I insert a variable for the sign, I get one implementation and an alias: https://godbolt.org/z/m-seT0
Does that mean that the performance will not regress?
There was a problem hiding this comment.
@tspiteri Oops, I simply made a typo in my test and that's why the result is different, duh :) It actually does produce the same IR, both with variable and without. So forget everything I said, this change is fine from the optimization perspective.
There was a problem hiding this comment.
@nikic While abs and wrapping_abs seem to be fine performance-wise, overflowing_abs from this PR actually does suffer from the problem discussed, so I will not forget everything you said! :) It's easily fixed by making it return (self.wrapping_abs(), self == Self::min_value()).
There was a problem hiding this comment.
And also, the code with the sign factored out is more readable. ;)
That said, I still think we should rather wait for CTFE to support conditionals than compromise code readability, and even the let sign version is still arcane magic. But that call is up to T-libs.
There was a problem hiding this comment.
No, we should definitely make as many of these small ops into const as soon as we can.
use `sign` variable in abs and wrapping_abs methods This also makes the code easier to understand by hinting at the significance of `self >> ($BITS - 1)`. Also, now `overflowing_abs` simply uses `wrapping_abs`, which is clearer and avoids a potential performance regression in the LLVM IR. This PR follows from the discussion from rust-lang#63786. r? @eddyb cc @nikic
Pkgsrc changes:
* Remove patch which no longer applies (but what about RPATH?)
* Adapt a few patches to changed files upstream.
Upstream changes:
Version 1.39.0 (2019-11-07)
===========================
Language
--------
- [You can now create `async` functions and blocks with `async fn`,
`async move {}`, and `async {}` respectively, and you can now call
`.await` on async expressions.][63209]
- [You can now use certain attributes on function, closure, and function
pointer parameters.][64010] These attributes include `cfg`, `cfg_attr`,
`allow`, `warn`, `deny`, `forbid` as well as inert helper attributes used
by procedural macro attributes applied to items. e.g.
```rust
fn len(
#[cfg(windows)] slice: &[u16],
#[cfg(not(windows))] slice: &[u8],
) -> usize {
slice.len()
}
```
- [You can now take shared references to bind-by-move patterns in the
`if` guards of `match` arms.][63118] e.g.
```rust
fn main() {
let array: Box<[u8; 4]> = Box::new([1, 2, 3, 4]);
match array {
nums
// ---- `nums` is bound by move.
if nums.iter().sum::<u8>() == 10
// ^------ `.iter()` implicitly takes a reference to `nums`.
=> {
drop(nums);
// ----------- Legal as `nums` was bound by move and so we have ownership.
}
_ => unreachable!(),
}
}
```
Compiler
--------
- [Added tier 3\* support for the `i686-unknown-uefi` target.][64334]
- [Added tier 3 support for the `sparc64-unknown-openbsd` target.][63595]
- [rustc will now trim code snippets in diagnostics to fit in your terminal.]
[63402] **Note** Cargo currently doesn't use this feature. Refer to
[cargo#7315][cargo/7315] to track this feature's progress.
- [You can now pass `--show-output` argument to test binaries to print the
output of successful tests.][62600]
\* Refer to Rust's [platform support page][forge-platform-support] for more
information on Rust's tiered platform support.
Libraries
---------
- [`Vec::new` and `String::new` are now `const` functions.][64028]
- [`LinkedList::new` is now a `const` function.][63684]
- [`str::len`, `[T]::len` and `str::as_bytes` are now `const` functions.][63770]
- [The `abs`, `wrapping_abs`, and `overflowing_abs` numeric functions are
now `const`.][63786]
Stabilized APIs
---------------
- [`Pin::into_inner`]
- [`Instant::checked_duration_since`]
- [`Instant::saturating_duration_since`]
Cargo
-----
- [You can now publish git dependencies if supplied with a `version`.]
[cargo/7237]
- [The `--all` flag has been renamed to `--workspace`.][cargo/7241] Using
`--all` is now deprecated.
Misc
----
- [You can now pass `-Clinker` to rustdoc to control the linker used
for compiling doctests.][63834]
Compatibility Notes
-------------------
- [Code that was previously accepted by the old borrow checker, but rejected by
the NLL borrow checker is now a hard error in Rust 2018.][63565] This was
previously a warning, and will also become a hard error in the Rust 2015
edition in the 1.40.0 release.
- [`rustdoc` now requires `rustc` to be installed and in the same directory to
run tests.][63827] This should improve performance when running a large
amount of doctests.
- [The `try!` macro will now issue a deprecation warning.][62672] It is
recommended to use the `?` operator instead.
- [`asinh(-0.0)` now correctly returns `-0.0`.][63698] Previously this
returned `0.0`.
[62600]: rust-lang/rust#62600
[62672]: rust-lang/rust#62672
[63118]: rust-lang/rust#63118
[63209]: rust-lang/rust#63209
[63402]: rust-lang/rust#63402
[63565]: rust-lang/rust#63565
[63595]: rust-lang/rust#63595
[63684]: rust-lang/rust#63684
[63698]: rust-lang/rust#63698
[63770]: rust-lang/rust#63770
[63786]: rust-lang/rust#63786
[63827]: rust-lang/rust#63827
[63834]: rust-lang/rust#63834
[63927]: rust-lang/rust#63927
[63933]: rust-lang/rust#63933
[63934]: rust-lang/rust#63934
[63938]: rust-lang/rust#63938
[63940]: rust-lang/rust#63940
[63941]: rust-lang/rust#63941
[63945]: rust-lang/rust#63945
[64010]: rust-lang/rust#64010
[64028]: rust-lang/rust#64028
[64334]: rust-lang/rust#64334
[cargo/7237]: rust-lang/cargo#7237
[cargo/7241]: rust-lang/cargo#7241
[cargo/7315]: rust-lang/cargo#7315
[`Pin::into_inner`]: https://doc.rust-lang.org/std/pin/struct.Pin.html#method.into_inner
[`Instant::checked_duration_since`]: https://doc.rust-lang.org/std/time/struct.Instant.html#method.checked_duration_since
[`Instant::saturating_duration_since`]: https://doc.rust-lang.org/std/time/struct.Instant.html#method.saturating_duration_since
Clean up const-hack PRs now that const if / match exist. Closes rust-lang#67627. Cleans up these merged PRs tagged with `const-hack`: - rust-lang#63810 - rust-lang#63786 - rust-lang#61635 - rust-lang#58044 reverting their contents to have the match or if expressions they originally contained. r? @oli-obk There's one more PR in those tagged with `const-hack` that originally wasn't merged (rust-lang#65107). Reading the thread, it looks like it was originally closed because the `const-hack` for the checked arithmetic non-negligibly hurt performance, and because there was no way to manipulate the returned Option at compile time anyway (with neither const if nor const match). Would you like me to add these changes to the changes from this PR here too, now that we have the necessary features?
This makes
abs,wrapping_absandoverflowing_absconst functions like #58044 makeswrapping_negandoverflowing_negconst functions.absis made const by returning(self ^ -1) - -1=!self + 1=-selffor negative numbers and(self ^ 0) - 0=selffor non-negative numbers. The subexpressionself >> ($BITS - 1)evaluates to-1for negative numbers and0otherwise. The subtraction overflows whenselfismin_value(), as we would be subtractingmax_value() - -1; this is whenabsshould overflow.wrapping_absandoverflowing_absmake use ofwrapping_subandoverflowing_subinstead of the subtraction operator.