Keep only the language item. This removes some indirection and makes
codegen worse for debug builds, but simplifies code significantly, which
is a good tradeoff to make, in my opinion.
Besides, the codegen can be improved even further with some constant
evaluation improvements that we expect to happen in the future.
This is necessary if we want to implement `[T]::align_to` and is more
useful in general.
This implementation effort has begun during the All Hands and represents
a month of my futile efforts to do any sort of maths. Luckily, I
found the very very nice Chris McDonald (cjm) on IRC who figured out the
core formulas for me! All the thanks for existence of this PR go to
them!
Anyway… Those formulas were mangled by yours truly into the arcane forms
you see here to squeeze out the best assembly possible on most of the
modern architectures (x86 and ARM were evaluated in practice). I mean,
just look at it: *one actual* modulo operation and everything else is
just the cheap single cycle ops! Admitedly, the naive solution might be
faster in some common scenarios, but this code absolutely butchers the
naive solution on the worst case scenario.
Alas, the result of this arcane magic also means that the code pretty
heavily relies on the preconditions holding true and breaking those
preconditions will unleash the UB-est of all UBs! So don’t.
Stabilize FusedIterator
FusedIterator is a marker trait that promises that the implementing
iterator continues to return `None` from `.next()` once it has returned
`None` once (and/or `.next_back()`, if implemented).
The effects of FusedIterator are already widely available through
`.fuse()`, but with stable `FusedIterator`, stable Rust users can
implement this trait for their iterators when appropriate.
Closes#35602
Backport LLVM fixes for a JumpThreading / assume intrinsic bug
This fixes the original cause of #48116 and restores the assume intrinsic that was removed as a workaround.
r? @alexcrichton
FusedIterator is a marker trait that promises that the implementing
iterator continues to return `None` from `.next()` once it has returned
`None` once (and/or `.next_back()`, if implemented).
The effects of FusedIterator are already widely available through
`.fuse()`, but with stable `FusedIterator`, stable Rust users can
implement this trait for their iterators when appropriate.
Use the slice length to hint the optimizer about iter.position result
Using the len of the iterator doesn't give the same result.
That's also why we can't generalize it to all TrustedLen iterators.
Problem demo: https://godbolt.org/g/MXg2ae
Fix demo: https://godbolt.org/g/P8q5aZ
Second attempt of #47333
Third attempt of #45501Fixes#45964
Standardize on "re-export" rather than "reexport"
While working on the book with our editors, it was brought to our attention that we're not consistent with when we use "re-export" versus "reexport". For the book, we've decided (with our editors) to go with "re-export"; in prose, I think that looks better. In code, I'm fine with "reexport".
However, the rustdoc generated section is currently "Reexports", so when we have a screenshot of generated documentation with the prose where we use "re-export", it's inconsistent.
It's too late to fix this for the book because we're using 1.21.0 for the output in the book, and it's really only one spot so it's not a huge deal, but I'd like to advocate for changing the documentation header so that a future edition of the book can be consistent.
The first commit here only changes the documentation section heading text and rustdoc documentation that references it. This is the commit that's most important to me.
The second commit changes error messages and associated tests to also be consistent with the use of re-export. This is the next most important commit to me, but I could be argued out of this one because then it won't match code like the `macro_reexports` feature name, which ostensibly should change to `macro_re_exports` to be most consistent but I didn't want to change code.
The last commit changes re-export anywhere else in prose: either in documentation comments or regular comments. This is least important as most of them aren't user-visible. Instances like these will likely sneak back in over time. I'm totally fine dropping this commit if anyone wants, but [the hobgoblins made me do it](http://www.bartleby.com/100/420.47.html) and it sets a good example.
r? @steveklabnik
Add slice::ExactChunks and ::ExactChunksMut iterators
These guarantee that always the requested slice size will be returned
and any leftoever elements at the end will be ignored. It allows llvm to
get rid of bounds checks in the code using the iterator.
This is inspired by the same iterators provided by ndarray.
Fixes https://github.com/rust-lang/rust/issues/47115
I'll add unit tests for all this if the general idea and behaviour makes sense for everybody.
Also see https://github.com/rust-lang/rust/issues/47115#issuecomment-354715511 for an example what this improves.
- Simplify nth() by making use of the fact that the slice is evenly
divisible by the chunk size, and calling next() instead of
duplicating it
- Call next_back() in last(), they are equivalent
- Implement ExactSizeIterator::is_empty()
These guarantee that always the requested slice size will be returned
and any leftoever elements at the end will be ignored. It allows llvm to
get rid of bounds checks in the code using the iterator.
This is inspired by the same iterators provided by ndarray.
See https://github.com/rust-lang/rust/issues/47115
Deprecate [T]::rotate in favor of [T]::rotate_{left,right}.
Background
==========
Slices currently have an **unstable** [`rotate`] method which rotates
elements in the slice to the _left_ N positions. [Here][tracking] is the
tracking issue for this unstable feature.
```rust
let mut a = ['a', 'b' ,'c', 'd', 'e', 'f'];
a.rotate(2);
assert_eq!(a, ['c', 'd', 'e', 'f', 'a', 'b']);
```
Proposal
========
Deprecate the [`rotate`] method and introduce `rotate_left` and
`rotate_right` methods.
```rust
let mut a = ['a', 'b' ,'c', 'd', 'e', 'f'];
a.rotate_left(2);
assert_eq!(a, ['c', 'd', 'e', 'f', 'a', 'b']);
```
```rust
let mut a = ['a', 'b' ,'c', 'd', 'e', 'f'];
a.rotate_right(2);
assert_eq!(a, ['e', 'f', 'a', 'b', 'c', 'd']);
```
Justification
=============
I used this method today for my first time and (probably because I’m a
naive westerner who reads LTR) was surprised when the docs mentioned that
elements get rotated in a left-ward direction. I was in a situation
where I needed to shift elements in a right-ward direction and had to
context switch from the main problem I was working on and think how much
to rotate left in order to accomplish the right-ward rotation I needed.
Ruby’s `Array.rotate` shifts left-ward, Python’s `deque.rotate` shifts
right-ward. Both of their implementations allow passing negative numbers
to shift in the opposite direction respectively. The current `rotate`
implementation takes an unsigned integer argument which doesn't allow
the negative number behavior.
Introducing `rotate_left` and `rotate_right` would:
- remove ambiguity about direction (alleviating need to read docs 😉)
- make it easier for people who need to rotate right
[`rotate`]: https://doc.rust-lang.org/std/primitive.slice.html#method.rotate
[tracking]: https://github.com/rust-lang/rust/issues/41891
