Optimize `escape_ascii` using a lookup table
Based upon my suggestion here: https://github.com/rust-lang/rust/pull/125340#issuecomment-2130441817
Effectively, we can take advantage of the fact that ASCII only needs 7 bits to make the eighth bit store whether the value should be escaped or not. This adds a 256-byte lookup table, but 256 bytes *should* be small enough that very few people will mind, according to my probably not incontrovertible opinion.
The generated assembly isn't clearly better (although has fewer branches), so, I decided to benchmark on three inputs: first on a random 200KiB, then on `/bin/cat`, then on `Cargo.toml` for this repo. In all cases, the generated code ran faster on my machine. (an old i7-8700)
But, if you want to try my benchmarking code for yourself:
<details><summary>Criterion code below. Replace <code>/home/ltdk/rustsrc</code> with the appropriate directory.</summary>
```rust
#![feature(ascii_char)]
#![feature(ascii_char_variants)]
#![feature(const_option)]
#![feature(let_chains)]
use core::ascii;
use core::ops::Range;
use criterion::{criterion_group, criterion_main, Criterion};
use rand::{thread_rng, Rng};
const HEX_DIGITS: [ascii::Char; 16] = *b"0123456789abcdef".as_ascii().unwrap();
#[inline]
const fn backslash<const N: usize>(a: ascii::Char) -> ([ascii::Char; N], Range<u8>) {
const { assert!(N >= 2) };
let mut output = [ascii::Char::Null; N];
output[0] = ascii::Char::ReverseSolidus;
output[1] = a;
(output, 0..2)
}
#[inline]
const fn hex_escape<const N: usize>(byte: u8) -> ([ascii::Char; N], Range<u8>) {
const { assert!(N >= 4) };
let mut output = [ascii::Char::Null; N];
let hi = HEX_DIGITS[(byte >> 4) as usize];
let lo = HEX_DIGITS[(byte & 0xf) as usize];
output[0] = ascii::Char::ReverseSolidus;
output[1] = ascii::Char::SmallX;
output[2] = hi;
output[3] = lo;
(output, 0..4)
}
#[inline]
const fn verbatim<const N: usize>(a: ascii::Char) -> ([ascii::Char; N], Range<u8>) {
const { assert!(N >= 1) };
let mut output = [ascii::Char::Null; N];
output[0] = a;
(output, 0..1)
}
/// Escapes an ASCII character.
///
/// Returns a buffer and the length of the escaped representation.
const fn escape_ascii_old<const N: usize>(byte: u8) -> ([ascii::Char; N], Range<u8>) {
const { assert!(N >= 4) };
match byte {
b'\t' => backslash(ascii::Char::SmallT),
b'\r' => backslash(ascii::Char::SmallR),
b'\n' => backslash(ascii::Char::SmallN),
b'\\' => backslash(ascii::Char::ReverseSolidus),
b'\'' => backslash(ascii::Char::Apostrophe),
b'\"' => backslash(ascii::Char::QuotationMark),
0x00..=0x1F => hex_escape(byte),
_ => match ascii::Char::from_u8(byte) {
Some(a) => verbatim(a),
None => hex_escape(byte),
},
}
}
/// Escapes an ASCII character.
///
/// Returns a buffer and the length of the escaped representation.
const fn escape_ascii_new<const N: usize>(byte: u8) -> ([ascii::Char; N], Range<u8>) {
/// Lookup table helps us determine how to display character.
///
/// Since ASCII characters will always be 7 bits, we can exploit this to store the 8th bit to
/// indicate whether the result is escaped or unescaped.
///
/// We additionally use 0x80 (escaped NUL character) to indicate hex-escaped bytes, since
/// escaped NUL will not occur.
const LOOKUP: [u8; 256] = {
let mut arr = [0; 256];
let mut idx = 0;
loop {
arr[idx as usize] = match idx {
// use 8th bit to indicate escaped
b'\t' => 0x80 | b't',
b'\r' => 0x80 | b'r',
b'\n' => 0x80 | b'n',
b'\\' => 0x80 | b'\\',
b'\'' => 0x80 | b'\'',
b'"' => 0x80 | b'"',
// use NUL to indicate hex-escaped
0x00..=0x1F | 0x7F..=0xFF => 0x80 | b'\0',
_ => idx,
};
if idx == 255 {
break;
}
idx += 1;
}
arr
};
let lookup = LOOKUP[byte as usize];
// 8th bit indicates escape
let lookup_escaped = lookup & 0x80 != 0;
// SAFETY: We explicitly mask out the eighth bit to get a 7-bit ASCII character.
let lookup_ascii = unsafe { ascii::Char::from_u8_unchecked(lookup & 0x7F) };
if lookup_escaped {
// NUL indicates hex-escaped
if matches!(lookup_ascii, ascii::Char::Null) {
hex_escape(byte)
} else {
backslash(lookup_ascii)
}
} else {
verbatim(lookup_ascii)
}
}
fn escape_bytes(bytes: &[u8], f: impl Fn(u8) -> ([ascii::Char; 4], Range<u8>)) -> Vec<ascii::Char> {
let mut vec = Vec::new();
for b in bytes {
let (buf, range) = f(*b);
vec.extend_from_slice(&buf[range.start as usize..range.end as usize]);
}
vec
}
pub fn criterion_benchmark(c: &mut Criterion) {
let mut group = c.benchmark_group("escape_ascii");
group.sample_size(1000);
let rand_200k = &mut [0; 200 * 1024];
thread_rng().fill(&mut rand_200k[..]);
let cat = include_bytes!("/bin/cat");
let cargo_toml = include_bytes!("/home/ltdk/rustsrc/Cargo.toml");
group.bench_function("old_rand", |b| {
b.iter(|| escape_bytes(rand_200k, escape_ascii_old));
});
group.bench_function("new_rand", |b| {
b.iter(|| escape_bytes(rand_200k, escape_ascii_new));
});
group.bench_function("old_bin", |b| {
b.iter(|| escape_bytes(cat, escape_ascii_old));
});
group.bench_function("new_bin", |b| {
b.iter(|| escape_bytes(cat, escape_ascii_new));
});
group.bench_function("old_cargo_toml", |b| {
b.iter(|| escape_bytes(cargo_toml, escape_ascii_old));
});
group.bench_function("new_cargo_toml", |b| {
b.iter(|| escape_bytes(cargo_toml, escape_ascii_new));
});
group.finish();
}
criterion_group!(benches, criterion_benchmark);
criterion_main!(benches);
```
</details>
My benchmark results:
```
escape_ascii/old_rand time: [1.6965 ms 1.7006 ms 1.7053 ms]
Found 22 outliers among 1000 measurements (2.20%)
4 (0.40%) high mild
18 (1.80%) high severe
escape_ascii/new_rand time: [1.6749 ms 1.6953 ms 1.7158 ms]
Found 38 outliers among 1000 measurements (3.80%)
38 (3.80%) high mild
escape_ascii/old_bin time: [224.59 µs 225.40 µs 226.33 µs]
Found 39 outliers among 1000 measurements (3.90%)
17 (1.70%) high mild
22 (2.20%) high severe
escape_ascii/new_bin time: [164.86 µs 165.63 µs 166.58 µs]
Found 107 outliers among 1000 measurements (10.70%)
43 (4.30%) high mild
64 (6.40%) high severe
escape_ascii/old_cargo_toml
time: [23.397 µs 23.699 µs 24.014 µs]
Found 204 outliers among 1000 measurements (20.40%)
21 (2.10%) high mild
183 (18.30%) high severe
escape_ascii/new_cargo_toml
time: [16.404 µs 16.438 µs 16.483 µs]
Found 88 outliers among 1000 measurements (8.80%)
56 (5.60%) high mild
32 (3.20%) high severe
```
Random: 1.7006ms => 1.6953ms (<1% speedup)
Binary: 225.40µs => 165.63µs (26% speedup)
Text: 23.699µs => 16.438µs (30% speedup)
Use throw intrinsic from stdarch in wasm libunwind
Tracking issue: #118168
This is a very belated followup to #121438; now that rust-lang/stdarch#1542 is merged, we can use the intrinsic exported from `core::arch` instead of defining it inline. I also cleaned up the cfgs a bit and added a more detailed comment.
remove const_cow_is_borrowed feature gate
The two functions guarded by this are still unstable, and there's no reason to require a separate feature gate for their const-ness -- we can just have `cow_is_borrowed` cover both kinds of stability.
Cc #65143
std: fix stdout-before-main
Fixes#130210.
Since #124881, `ReentrantLock` uses `ThreadId` to identify threads. This has the unfortunate consequence of breaking uses of `Stdout` before main: Locking the `ReentrantLock` that synchronizes the output will initialize the thread ID before the handle for the main thread is set in `rt::init`. But since that would overwrite the current thread ID, `thread::set_current` triggers an abort.
This PR fixes the problem by using the already initialized thread ID for constructing the main thread handle and allowing `set_current` calls that do not change the thread's ID.
Stabilize const `ptr::write*` and `mem::replace`
Since `const_mut_refs` and `const_refs_to_cell` have been stabilized, we may now also stabilize the ability to write to places during const evaluation inside our library API. So, we now propose the `const fn` version of `ptr::write` and its variants. This allows us to also stabilize `mem::replace` and `ptr::replace`.
- const `mem::replace`: https://github.com/rust-lang/rust/issues/83164#issuecomment-2338660862
- const `ptr::write{,_bytes,_unaligned}`: https://github.com/rust-lang/rust/issues/86302#issuecomment-2330275266
Their implementation requires an additional internal stabilization of `const_intrinsic_forget`, which is required for `*::write*` and thus `*::replace`. Thus we const-stabilize the internal intrinsics `forget`, `write_bytes`, and `write_via_move`.
Fixes#130210.
Since #124881, `ReentrantLock` uses `ThreadId` to identify threads. This has the unfortunate consequence of breaking uses of `Stdout` before main: Locking the `ReentrantLock` that synchronizes the output will initialize the thread ID before the handle for the main thread is set in `rt::init`. But since that would overwrite the current thread ID, `thread::set_current` triggers an abort.
This PR fixes the problem by using the already initialized thread ID for constructing the main thread handle and allowing `set_current` calls that do not change the thread's ID.
Migrate lib's `&Option<T>` into `Option<&T>`
Trying out my new lint https://github.com/rust-lang/rust-clippy/pull/13336 - according to the [video](https://www.youtube.com/watch?v=6c7pZYP_iIE), this could lead to some performance and memory optimizations.
Basic thoughts expressed in the video that seem to make sense:
* `&Option<T>` in an API breaks encapsulation:
* caller must own T and move it into an Option to call with it
* if returned, the owner must store it as Option<T> internally in order to return it
* Performance is subject to compiler optimization, but at the basics, `&Option<T>` points to memory that has `presence` flag + value, whereas `Option<&T>` by specification is always optimized to a single pointer.
intrinsics fmuladdf{32,64}: expose llvm.fmuladd.* semantics
Add intrinsics `fmuladd{f32,f64}`. This computes `(a * b) + c`, to be fused if the code generator determines that (i) the target instruction set has support for a fused operation, and (ii) that the fused operation is more efficient than the equivalent, separate pair of `mul` and `add` instructions.
https://llvm.org/docs/LangRef.html#llvm-fmuladd-intrinsic
The codegen_cranelift uses the `fma` function from libc, which is a correct implementation, but without the desired performance semantic. I think this requires an update to cranelift to expose a suitable instruction in its IR.
I have not tested with codegen_gcc, but it should behave the same way (using `fma` from libc).
---
This topic has been discussed a few times on Zulip and was suggested, for example, by `@workingjubilee` in [Effect of fma disabled](https://rust-lang.zulipchat.com/#narrow/stream/122651-general/topic/Effect.20of.20fma.20disabled/near/274179331).
Stabilise `const_char_encode_utf8`.
Closes: #130512
This PR stabilises the `const_char_encode_utf8` feature gate (i.e. support for `char::encode_utf8` in const scenarios).
Note that the linked tracking issue is currently awaiting FCP.
Port sort-research-rs test suite to Rust stdlib tests
This PR is a followup to https://github.com/rust-lang/rust/pull/124032. It replaces the tests that test the various sort functions in the standard library with a test-suite developed as part of https://github.com/Voultapher/sort-research-rs. The current tests suffer a couple of problems:
- They don't cover important real world patterns that the implementations take advantage of and execute special code for.
- The input lengths tested miss out on code paths. For example, important safety property tests never reach the quicksort part of the implementation.
- The miri side is often limited to `len <= 20` which means it very thoroughly tests the insertion sort, which accounts for 19 out of 1.5k LoC.
- They are split into to core and alloc, causing code duplication and uneven coverage.
- ~~The randomness is tied to a caller location, wasting the space exploration capabilities of randomized testing.~~ The randomness is not repeatable, as it relies on `std:#️⃣:RandomState::new().build_hasher()`.
Most of these issues existed before https://github.com/rust-lang/rust/pull/124032, but they are intensified by it. One thing that is new and requires additional testing, is that the new sort implementations specialize based on type properties. For example `Freeze` and non `Freeze` execute different code paths.
Effectively there are three dimensions that matter:
- Input type
- Input length
- Input pattern
The ported test-suite tests various properties along all three dimensions, greatly improving test coverage. It side-steps the miri issue by preferring sampled approaches. For example the test that checks if after a panic the set of elements is still the original one, doesn't do so for every single possible panic opportunity but rather it picks one at random, and performs this test across a range of input length, which varies the panic point across them. This allows regular execution to easily test inputs of length 10k, and miri execution up to 100 which covers significantly more code. The randomness used is tied to a fixed - but random per process execution - seed. This allows for fully repeatable tests and fuzzer like exploration across multiple runs.
Structure wise, the tests are previously found in the core integration tests for `sort_unstable` and alloc unit tests for `sort`. The new test-suite was developed to be a purely black-box approach, which makes integration testing the better place, because it can't accidentally rely on internal access. Because unwinding support is required the tests can't be in core, even if the implementation is, so they are now part of the alloc integration tests. Are there architectures that can only build and test core and not alloc? If so, do such platforms require sort testing? For what it's worth the current implementation state passes miri `--target mips64-unknown-linux-gnuabi64` which is big endian.
The test-suite also contains tests for properties that were and are given by the current and previous implementations, and likely relied upon by users but weren't tested. For example `self_cmp` tests that the two parameters `a` and `b` passed into the comparison function are never references to the same object, which if the user is sorting for example a `&mut [Mutex<i32>]` could lead to a deadlock.
Instead of using the hashed caller location as rand seed, it uses seconds since unix epoch / 10, which given timestamps in the CI should be reasonably easy to reproduce, but also allows fuzzer like space exploration.
---
Test run-time changes:
Setup:
```
Linux 6.10
rustc 1.83.0-nightly (f79a912d9 2024-09-18)
AMD Ryzen 9 5900X 12-Core Processor (Zen 3 micro-architecture)
CPU boost enabled.
```
master: e9df22f
Before core integration tests:
```
$ LD_LIBRARY_PATH=build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/ hyperfine build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/coretests-219cbd0308a49e2f
Time (mean ± σ): 869.6 ms ± 21.1 ms [User: 1327.6 ms, System: 95.1 ms]
Range (min … max): 845.4 ms … 917.0 ms 10 runs
# MIRIFLAGS="-Zmiri-disable-isolation" to get real time
$ MIRIFLAGS="-Zmiri-disable-isolation" ./x.py miri library/core
finished in 738.44s
```
After core integration tests:
```
$ LD_LIBRARY_PATH=build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/ hyperfine build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/coretests-219cbd0308a49e2f
Time (mean ± σ): 865.1 ms ± 14.7 ms [User: 1283.5 ms, System: 88.4 ms]
Range (min … max): 836.2 ms … 885.7 ms 10 runs
$ MIRIFLAGS="-Zmiri-disable-isolation" ./x.py miri library/core
finished in 752.35s
```
Before alloc unit tests:
```
LD_LIBRARY_PATH=build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/ hyperfine build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/alloc-19c15e6e8565aa54
Time (mean ± σ): 295.0 ms ± 9.9 ms [User: 719.6 ms, System: 35.3 ms]
Range (min … max): 284.9 ms … 319.3 ms 10 runs
$ MIRIFLAGS="-Zmiri-disable-isolation" ./x.py miri library/alloc
finished in 322.75s
```
After alloc unit tests:
```
LD_LIBRARY_PATH=build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/ hyperfine build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/alloc-19c15e6e8565aa54
Time (mean ± σ): 97.4 ms ± 4.1 ms [User: 297.7 ms, System: 28.6 ms]
Range (min … max): 92.3 ms … 109.2 ms 27 runs
$ MIRIFLAGS="-Zmiri-disable-isolation" ./x.py miri library/alloc
finished in 309.18s
```
Before alloc integration tests:
```
$ LD_LIBRARY_PATH=build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/ hyperfine build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/alloctests-439e7300c61a8046
Time (mean ± σ): 103.2 ms ± 1.7 ms [User: 135.7 ms, System: 39.4 ms]
Range (min … max): 99.7 ms … 107.3 ms 28 runs
$ MIRIFLAGS="-Zmiri-disable-isolation" ./x.py miri library/alloc
finished in 231.35s
```
After alloc integration tests:
```
$ LD_LIBRARY_PATH=build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/ hyperfine build/x86_64-unknown-linux-gnu/stage0-std/x86_64-unknown-linux-gnu/release/deps/alloctests-439e7300c61a8046
Time (mean ± σ): 379.8 ms ± 4.7 ms [User: 4620.5 ms, System: 1157.2 ms]
Range (min … max): 373.6 ms … 386.9 ms 10 runs
$ MIRIFLAGS="-Zmiri-disable-isolation" ./x.py miri library/alloc
finished in 449.24s
```
In my opinion the results don't change iterative library development or CI execution in meaningful ways. For example currently the library doc-tests take ~66s and incremental compilation takes 10+ seconds. However I only have limited knowledge of the various local development workflows that exist, and might be missing one that is significantly impacted by this change.
Add intrinsics `fmuladd{f16,f32,f64,f128}`. This computes `(a * b) +
c`, to be fused if the code generator determines that (i) the target
instruction set has support for a fused operation, and (ii) that the
fused operation is more efficient than the equivalent, separate pair
of `mul` and `add` instructions.
https://llvm.org/docs/LangRef.html#llvm-fmuladd-intrinsic
MIRI support is included for f32 and f64.
The codegen_cranelift uses the `fma` function from libc, which is a
correct implementation, but without the desired performance semantic. I
think this requires an update to cranelift to expose a suitable
instruction in its IR.
I have not tested with codegen_gcc, but it should behave the same
way (using `fma` from libc).
rustc_target: Add sme-b16b16 as an explicit aarch64 target feature
LLVM 20 split out what used to be called b16b16 and correspond to aarch64
FEAT_SVE_B16B16 into sve-b16b16 and sme-b16b16.
Add sme-b16b16 as an explicit feature and update the codegen accordingly.
Resolves https://github.com/rust-lang/rust/pull/129894.
Stabilize const `{slice,array}::from_mut`
This PR stabilizes the following APIs as const stable as of rust `1.83`:
```rs
// core::array
pub const fn from_mut<T>(s: &mut T) -> &mut [T; 1];
// core::slice
pub const fn from_mut<T>(s: &mut T) -> &mut [T];
```
This is made possible by `const_mut_refs` being stabilized (yay).
Tracking issue: #90206
LLVM 20 split out what used to be called b16b16 and correspond to aarch64
FEAT_SVE_B16B16 into sve-b16b16 and sme-b16b16.
Add sme-b16b16 as an explicit feature and update the codegen accordingly.
Decouple WASIp2 sockets from WasiFd
This is a follow up to #129638, decoupling WASIp2's socket implementation from WASIp1's `WasiFd` as discussed with `@alexcrichton.`
Quite a few trait implementations in `std::os::fd` rely on the fact that there is an additional layer of abstraction between `Socket` and `OwnedFd`. I thus had to add a thin `WasiSocket` wrapper struct that just "forwards" to `OwnedFd`. Alternatively, I could have added a lot of conditional compilation to `std::os::fd`, which feels even worse.
Since `WasiFd::sock_accept` is no longer accessible from `TcpListener` and since WASIp2 has proper support for accepting sockets through `Socket::accept`, the `std::os::wasi::net` module has been removed from WASIp2, which only contains a single `TcpListenerExt` trait with a `sock_accept` method as well as an implementation for `TcpListener`. Let me know if this is an acceptable solution.
