Rollup of 9 pull requests
Successful merges:
- #48658 (Add a generic CAS loop to std::sync::Atomic*)
- #49253 (Take the original extra-filename passed to a crate into account when resolving it as a dependency)
- #49345 (RFC 2008: Finishing Touches)
- #49432 (Flush executables to disk after linkage)
- #49496 (Add more vec![... ; n] optimizations)
- #49563 (add a dist builder to build rust-std components for the THUMB targets)
- #49654 (Host compiler documentation: Include private items)
- #49667 (Add more features to rust_2018_preview)
- #49674 (ci: Remove x86_64-gnu-incremental builder)
Failed merges:
Add more vec![... ; n] optimizations
vec![0; n], via implementations of SpecFromElem, has an optimization that uses with_capacity_zeroed instead of with_capacity, which will use calloc instead of malloc, and avoid an extra memset.
This PR adds the same optimization for ptr::null, ptr::null_mut, and None, when their in-memory representation is zeroes.
Introduce Vec::resize_with method (see #41758)
In #41758, the libs team decided they preferred `Vec::resize_with` over `Vec::resize_default()`. Here is an implementation to get this moving forward.
I don't know what the removal process for `Vec::resize_default()` should be, so I've left it in place for now. Would be happy to follow up with its removal.
Add #[must_use] to a few standard library methods
Chosen to start a precedent of using it on ones that are potentially-expensive and where using it for side effects is particularly discouraged.
Discuss :)
```rust
warning: unused return value of `std::iter::Iterator::collect` which must be used: if you really need to exhaust the iterator, consider `.for_each(drop)` instead
--> $DIR/fn_must_use_stdlib.rs:19:5
|
LL | "1 2 3".split_whitespace().collect::<Vec<_>>();
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
warning: unused return value of `std::borrow::ToOwned::to_owned` which must be used: cloning is often expensive and is not expected to have side effects
--> $DIR/fn_must_use_stdlib.rs:21:5
|
LL | "hello".to_owned();
| ^^^^^^^^^^^^^^^^^^^
warning: unused return value of `std::clone::Clone::clone` which must be used: cloning is often expensive and is not expected to have side effects
--> $DIR/fn_must_use_stdlib.rs:23:5
|
LL | String::from("world").clone();
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
```
cc https://github.com/rust-lang/rust/issues/48926
94d1970bba moved the alloc::allocator
module to core::heap, moving e.g. Alloc and Layout out of the alloc
crate. While alloc::heap reexports them, it's better to use them from
where they really come from.
vec![0; n], via implementations of SpecFromElem, has an optimization
that uses with_capacity_zeroed instead of with_capacity, which will use
calloc instead of malloc, and avoid an extra memset.
This adds the same optimization for vec![ptr::null(); n] and
vec![ptr::null_mut(); n], assuming their bit value is 0 (which is true
on all currently supported platforms).
This does so by adding an intermediate trait IsZero, which looks very
much like nonzero::Zeroable, but that one is on the way out, and doesn't
apply to pointers anyways.
Adding such a trait allows to avoid repeating the logic using
with_capacity_zeroed or with_capacity, or making the macro more complex
to support generics.
Chosen to start a precedent of using it on ones that are potentially-expensive and where using it for side effects is particularly discouraged.
Discuss :)
Use f{32,64}::to_bits for is_zero test in vec::SpecFromElem
vec::SpecFromElem provides an optimization to use calloc to fill a Vec
when the element given to fill the Vec is represented by 0.
For floats, the test for that currently used is `x == 0. &&
x.is_sign_positive()`. When compiled in a standalone function, rustc
generates the following assembly:
```
xorps xmm1, xmm1
ucomisd xmm0, xmm1
setnp al
sete cl
and cl, al
movq rax, xmm0
test rax, rax
setns al
and al, cl
ret
```
A simpler test telling us whether the value is represented by 0, is
`x.to_bits() == 0`, which rustc compiles to:
```
movq rax, xmm0
test rax, rax
sete al
ret
```
Not that the test is hot in any way, but it also makes it clearer what
the intent in the rust code is.
vec::SpecFromElem provides an optimization to use calloc to fill a Vec
when the element given to fill the Vec is represented by 0.
For floats, the test for that currently used is `x == 0. &&
x.is_sign_positive()`. When compiled in a standalone function, rustc
generates the following assembly:
```
xorps xmm1, xmm1
ucomisd xmm0, xmm1
setnp al
sete cl
and cl, al
movq rax, xmm0
test rax, rax
setns al
and al, cl
ret
```
A simpler test telling us whether the value is represented by 0, is
`x.to_bits() == 0`, which rustc compiles to:
```
movq rax, xmm0
test rax, rax
sete al
ret
```
Not that the test is hot in any way, but it also makes it clearer what
the intent in the rust code is.
Add is_whitespace and is_alphanumeric to str.
The other methods from `UnicodeStr` are already stable inherent
methods on str, but these have not been included.
r? @SimonSapin
Add slice::sort_by_cached_key as a memoised sort_by_key
At present, `slice::sort_by_key` calls its key function twice for each comparison that is made. When the key function is expensive (which can often be the case when `sort_by_key` is chosen over `sort_by`), this can lead to very suboptimal behaviour.
To address this, I've introduced a new slice method, `sort_by_cached_key`, which has identical semantic behaviour to `sort_by_key`, except that it guarantees the key function will only be called once per element.
Where there are `n` elements and the key function is `O(m)`:
- `slice::sort_by_cached_key` time complexity is `O(m n log m n)`, compared to `slice::sort_by_key`'s `O(m n + n log n)`.
- `slice::sort_by_cached_key` space complexity remains at `O(n + m)`. (Technically, it now reserves a slice of size `n`, whereas before it reserved a slice of size `n/2`.)
`slice::sort_unstable_by_key` has not been given an analogue, as it is important that unstable sorts are in-place, which is not a property that is guaranteed here. However, this also means that `slice::sort_unstable_by_key` is likely to be slower than `slice::sort_by_cached_key` when the key function does not have negligible complexity. We might want to explore this trade-off further in the future.
Benchmarks (for a vector of 100 `i32`s):
```
# Lexicographic: `|x| x.to_string()`
test bench_sort_by_key ... bench: 112,638 ns/iter (+/- 19,563)
test bench_sort_by_cached_key ... bench: 15,038 ns/iter (+/- 4,814)
# Identity: `|x| *x`
test bench_sort_by_key ... bench: 1,346 ns/iter (+/- 238)
test bench_sort_by_cached_key ... bench: 1,839 ns/iter (+/- 765)
# Power: `|x| x.pow(31)`
test bench_sort_by_key ... bench: 3,624 ns/iter (+/- 738)
test bench_sort_by_cached_key ... bench: 1,997 ns/iter (+/- 311)
# Abs: `|x| x.abs()`
test bench_sort_by_key ... bench: 1,546 ns/iter (+/- 174)
test bench_sort_by_cached_key ... bench: 1,668 ns/iter (+/- 790)
```
(So it seems functions that are single operations do perform slightly worse with this method, but for pretty much any more complex key, you're better off with this optimisation.)
I've definitely found myself using expensive keys in the past and wishing this optimisation was made (e.g. for https://github.com/rust-lang/rust/pull/47415). This feels like both desirable and expected behaviour, at the small cost of slightly more stack allocation and minute degradation in performance for extremely trivial keys.
Resolves#34447.
Simon Sapin