This commit is an implementation of [RFC 1184][rfc] which tweaks the behavior of
the `#![no_std]` attribute and adds a new `#![no_core]` attribute. The
`#![no_std]` attribute now injects `extern crate core` at the top of the crate
as well as the libcore prelude into all modules (in the same manner as the
standard library's prelude). The `#![no_core]` attribute disables both std and
core injection.
[rfc]: https://github.com/rust-lang/rfcs/pull/1184
The following APIs were all marked with a `#[stable]` tag:
* process::Child::id
* error::Error::is
* error::Error::downcast
* error::Error::downcast_ref
* error::Error::downcast_mut
* io::Error::get_ref
* io::Error::get_mut
* io::Error::into_inner
* hash::Hash::hash_slice
* hash::Hasher::write_{i,u}{8,16,32,64,size}
The following APIs were all marked with a `#[stable]` tag:
* process::Child::id
* error::Error::is
* error::Error::downcast
* error::Error::downcast_ref
* error::Error::downcast_mut
* io::Error::get_ref
* io::Error::get_mut
* io::Error::into_inner
* hash::Hash::hash_slice
* hash::Hasher::write_{i,u}{8,16,32,64,size}
Improve siphash performance for longer data
Use `ptr::copy_nonoverlapping` (aka memcpy) to load an u64 from the
byte stream. This is correct for any alignment, and the compiler will
use the appropriate instruction to load the data.
Also contains small tweaks that should benefit hashing short data too,
both the commit that removes a variable and the autovectorization of
the hash state initialization (in SipHash::reset).
Benchmarks show that hashing longer data benefits for the improved word loading.
Before (using benchmarks from the first commit in the PR):
The before benchmark is a bit noisy.
```
test hash::sip::bench_bytes_4 ... bench: 41 ns/iter (+/- 0) = 97 MB/s
test hash::sip::bench_bytes_7 ... bench: 49 ns/iter (+/- 2) = 142 MB/s
test hash::sip::bench_bytes_8 ... bench: 42 ns/iter (+/- 4) = 190 MB/s
test hash::sip::bench_bytes_a_16 ... bench: 57 ns/iter (+/- 14) = 280 MB/s
test hash::sip::bench_bytes_b_32 ... bench: 85 ns/iter (+/- 74) = 376 MB/s
test hash::sip::bench_bytes_c_128 ... bench: 278 ns/iter (+/- 33) = 460 MB/s
test hash::sip::bench_long_str ... bench: 825 ns/iter (+/- 103)
test hash::sip::bench_str_of_8_bytes ... bench: 151 ns/iter (+/- 66)
test hash::sip::bench_str_over_8_bytes ... bench: 59 ns/iter (+/- 3)
test hash::sip::bench_str_under_8_bytes ... bench: 47 ns/iter (+/- 56)
test hash::sip::bench_u32 ... bench: 39 ns/iter (+/- 93) = 205 MB/s
test hash::sip::bench_u32_keyed ... bench: 40 ns/iter (+/- 88) = 200 MB/s
test hash::sip::bench_u64 ... bench: 54 ns/iter (+/- 96) = 148 MB/s
```
After:
```
test hash::sip::bench_bytes_4 ... bench: 41 ns/iter (+/- 3) = 97 MB/s
test hash::sip::bench_bytes_7 ... bench: 48 ns/iter (+/- 0) = 145 MB/s
test hash::sip::bench_bytes_8 ... bench: 35 ns/iter (+/- 1) = 228 MB/s
test hash::sip::bench_bytes_a_16 ... bench: 45 ns/iter (+/- 1) = 355 MB/s
test hash::sip::bench_bytes_b_32 ... bench: 60 ns/iter (+/- 0) = 533 MB/s
test hash::sip::bench_bytes_c_128 ... bench: 161 ns/iter (+/- 5) = 795 MB/s
test hash::sip::bench_long_str ... bench: 514 ns/iter (+/- 5)
test hash::sip::bench_str_of_8_bytes ... bench: 44 ns/iter (+/- 0)
test hash::sip::bench_str_over_8_bytes ... bench: 51 ns/iter (+/- 0)
test hash::sip::bench_str_under_8_bytes ... bench: 52 ns/iter (+/- 6)
test hash::sip::bench_u32 ... bench: 40 ns/iter (+/- 2) = 200 MB/s
test hash::sip::bench_u32_keyed ... bench: 39 ns/iter (+/- 1) = 205 MB/s
test hash::sip::bench_u64 ... bench: 36 ns/iter (+/- 1) = 222 MB/s
```
Many of these have long since reached their stage of being obsolete, so this
commit starts the removal process for all of them. The unstable features that
were deprecated are:
* cmp_partial
* fs_time
* hash_default
* int_slice
* iter_min_max
* iter_reset_fuse
* iter_to_vec
* map_in_place
* move_from
* owned_ascii_ext
* page_size
* read_and_zero
* scan_state
* slice_chars
* slice_position_elem
* subslice_offset
If they are ordered v0, v2, v1, v3, the compiler can find just a few
simd optimizations itself.
The new optimization I could observe on x86-64 was using 128 bit
registers for the v = key ^ constant operations in new / reset.
Use `ptr::copy_nonoverlapping` (aka memcpy) to load an u64 from the
byte stream. This is correct for any alignment, and the compiler will
use the appropriate instruction to load the data.
Use unchecked indexing.
This results in a large improvement of throughput (hashed bytes
/ second) for long data. Maximum improvement benches at a 70% increase
in throughput for large values (> 256 bytes) but already values of 16
bytes or larger improve.
Introducing unchecked indexing is motivated to reach as good throughput
as possible. Using ptr::copy_nonoverlapping without unchecked indexing
would land the improvement some 20-30 pct units lower.
We use a debug assertion so that the test suite checks our use of
unchecked indexing.
This commit shards the broad `core` feature of the libcore library into finer
grained features. This split groups together similar APIs and enables tracking
each API separately, giving a better sense of where each feature is within the
stabilization process.
A few minor APIs were deprecated along the way:
* Iterator::reverse_in_place
* marker::NoCopy
Since the hashmap and its hasher are implemented in different crates, we
currently can't benefit from inlining, which means that especially for
small, fixed size keys, there is a huge overhead in hash calculations,
because the compiler can't apply optimizations that only apply for these
keys.
Fixes the brainfuck benchmark in #24014.
This is an implementation of [RFC 1030][rfc] which adds these traits to the
prelude and additionally removes all inherent `into_iter` methods on collections
in favor of the trait implementation (which is now accessible by default).
[rfc]: https://github.com/rust-lang/rfcs/pull/1030
This is technically a breaking change due to the prelude additions and removal
of inherent methods, but it is expected that essentially no code breaks in
practice.
[breaking-change]
Closes#24538
This is an implementation of [RFC 1030][rfc] which adds these traits to the
prelude and additionally removes all inherent `into_iter` methods on collections
in favor of the trait implementation (which is now accessible by default).
[rfc]: https://github.com/rust-lang/rfcs/pull/1030
This is technically a breaking change due to the prelude additions and removal
of inherent methods, but it is expected that essentially no code breaks in
practice.
[breaking-change]
Closes#24538
These constants are small and can fit even in `u8`, but semantically they have type `usize` because they denote sizes and are almost always used in `usize` context. The change of their type to `u32` during the integer audit led only to the large amount of `as usize` noise (see the second commit, which removes this noise).
This is a minor [breaking-change] to an unstable interface.
r? @aturon
This commit cleans out a large amount of deprecated APIs from the standard
library and some of the facade crates as well, updating all users in the
compiler and in tests as it goes along.
This commit stabilizes the `std::num` module:
* The `Int` and `Float` traits are deprecated in favor of (1) the
newly-added inherent methods and (2) the generic traits available in
rust-lang/num.
* The `Zero` and `One` traits are reintroduced in `std::num`, which
together with various other traits allow you to recover the most
common forms of generic programming.
* The `FromStrRadix` trait, and associated free function, is deprecated
in favor of inherent implementations.
* A wide range of methods and constants for both integers and floating
point numbers are now `#[stable]`, having been adjusted for integer
guidelines.
* `is_positive` and `is_negative` are renamed to `is_sign_positive` and
`is_sign_negative`, in order to address #22985
* The `Wrapping` type is moved to `std::num` and stabilized;
`WrappingOps` is deprecated in favor of inherent methods on the
integer types, and direct implementation of operations on
`Wrapping<X>` for each concrete integer type `X`.
Closes#22985Closes#21069
[breaking-change]
This permits all coercions to be performed in casts, but adds lints to warn in those cases.
Part of this patch moves cast checking to a later stage of type checking. We acquire obligations to check casts as part of type checking where we previously checked them. Once we have type checked a function or module, then we check any cast obligations which have been acquired. That means we have more type information available to check casts (this was crucial to making coercions work properly in place of some casts), but it means that casts cannot feed input into type inference.
[breaking change]
* Adds two new lints for trivial casts and trivial numeric casts, these are warn by default, but can cause errors if you build with warnings as errors. Previously, trivial numeric casts and casts to trait objects were allowed.
* The unused casts lint has gone.
* Interactions between casting and type inference have changed in subtle ways. Two ways this might manifest are:
- You may need to 'direct' casts more with extra type information, for example, in some cases where `foo as _ as T` succeeded, you may now need to specify the type for `_`
- Casts do not influence inference of integer types. E.g., the following used to type check:
```
let x = 42;
let y = &x as *const u32;
```
Because the cast would inform inference that `x` must have type `u32`. This no longer applies and the compiler will fallback to `i32` for `x` and thus there will be a type error in the cast. The solution is to add more type information:
```
let x: u32 = 42;
let y = &x as *const u32;
```
This commit enables writing a stable implementation of the `Hasher` trait as
well as actually calculating the hash of a vlaue in a stable fashion. The
signature is stabilized as-is.
This commit enables writing a stable implementation of the `Hasher` trait as
well as actually calculating the hash of a vlaue in a stable fashion. The
signature is stabilized as-is.
Many of the core rust libraries have places that rely on integer
wrapping behaviour. These places have been altered to use the wrapping_*
methods:
* core:#️⃣:sip - A number of macros
* core::str - The `maximal_suffix` method in `TwoWaySearcher`
* rustc::util::nodemap - Implementation of FnvHash
* rustc_back::sha2 - A number of macros and other places
* rand::isaac - Isaac64Rng, changed to use the Wrapping helper type
Some places had "benign" underflow. This is when underflow or overflow
occurs, but the unspecified value is not used due to other conditions.
* collections::bit::Bitv - underflow when `self.nbits` is zero.
* collections:#️⃣:{map,table} - Underflow when searching an empty
table. Did cause undefined behaviour in this case due to an
out-of-bounds ptr::offset based on the underflowed index. However the
resulting pointers would never be read from.
* syntax::ext::deriving::encodable - Underflow when calculating the
index of the last field in a variant with no fields.
These cases were altered to avoid the underflow, often by moving the
underflowing operation to a place where underflow could not happen.
There was one case that relied on the fact that unsigned arithmetic and
two's complement arithmetic are identical with wrapping semantics. This
was changed to use the wrapping_* methods.
Finally, the calculation of variant discriminants could overflow if the
preceeding discriminant was `U64_MAX`. The logic in `rustc::middle::ty`
for this was altered to avoid the overflow completely, while the
remaining places were changed to use wrapping methods. This is because
`rustc::middle::ty::enum_variants` now throws an error when the
calculated discriminant value overflows a `u64`.
This behaviour can be triggered by the following code:
```
enum Foo {
A = U64_MAX,
B
}
```
This commit also implements the remaining integer operators for
Wrapped<T>.
* count_ones/zeros, trailing_ones/zeros return u32, not usize
* rotate_left/right take u32, not usize
* RADIX, MANTISSA_DIGITS, DIGITS, BITS, BYTES are u32, not usize
Doesn't touch pow because there's another PR for it.
[breaking-change]
This commit stabilizes `std::borrow`, making the following modifications
to catch up the API with language changes:
* It renames `BorrowFrom` to `Borrow`, as was originally intended (but
blocked for technical reasons), and reorders the parameters
accordingly.
* It moves the type parameter of `ToOwned` to an associated type. This
is somewhat less flexible, in that each borrowed type must have a
unique owned type, but leads to a significant simplification for
`Cow`. Flexibility can be regained by using newtyped slices, which is
advisable for other reasons anyway.
* It removes the owned type parameter from `Cow`, making the type much
less verbose.
* Deprecates the `is_owned` and `is_borrowed` predicates in favor of
direct matching.
The above API changes are relatively minor; the basic functionality
remains the same, and essentially the whole module is now marked
`#[stable]`.
[breaking-change]
This commit is an implementation of [RFC 823][rfc] which is another pass over
the `std::hash` module for stabilization. The contents of the module were not
entirely marked stable, but some portions which remained quite similar to the
previous incarnation are now marked `#[stable]`. Specifically:
[rfc]: https://github.com/rust-lang/rfcs/blob/master/text/0823-hash-simplification.md
* `std::hash` is now stable (the name)
* `Hash` is now stable
* `Hash::hash` is now stable
* `Hasher` is now stable
* `SipHasher` is now stable
* `SipHasher::new` and `new_with_keys` are now stable
* `Hasher for SipHasher` is now stable
* Many `Hash` implementations are now stable
All other portions of the `hash` module remain `#[unstable]` as they are less
commonly used and were recently redesigned.
This commit is a breaking change due to the modifications to the `std::hash` API
and more details can be found on the [RFC][rfc].
Closes#22467
[breaking-change]
This was particularly helpful in the time just after OIBIT's
implementation to make sure things that were supposed to be Copy
continued to be, but it's now creates a lot of noise for types that
intentionally don't want to be Copy.
r? @alexcrichton
This was particularly helpful in the time just after OIBIT's
implementation to make sure things that were supposed to be Copy
continued to be, but it's now creates a lot of noise for types that
intentionally don't want to be Copy.
