Add the core execution types, the borrowed context, and the seeded u64 fx hasher

This commit is contained in:
Dennis Kobert
2026-07-31 13:11:36 +00:00
parent eb50be844b
commit 1fccb112cb
14 changed files with 2263 additions and 61 deletions

View File

@@ -68,6 +68,7 @@ impl_via_hash! {
#[cfg(feature = "std")]
impl_via_hash! {
String,
core::time::Duration,
}
impl<'a> CacheHash for std::borrow::Cow<'a, str> {
@@ -235,3 +236,142 @@ impl_tuple!(A, B, C);
impl_tuple!(A, B, C, D);
impl_tuple!(A, B, C, D, E);
impl_tuple!(A, B, C, D, E, F);
/// rustc-hash's polynomial hash with the state pinned to u64, so keys match across native and wasm targets.
/// The state starts at a nonzero seed: zero-initialized fx absorbs leading zero words, which produced
/// a real wrong-value memo hit in the prototype.
#[derive(Clone)]
pub struct FxHasher64 {
hash: u64,
}
const K: u64 = 0xf1357aea2e62a9c5;
const SEED: u64 = 0x517cc1b727220a95;
const SEED1: u64 = 0x243f6a8885a308d3;
const SEED2: u64 = 0x13198a2e03707344;
const PREVENT_TRIVIAL_ZERO_COLLAPSE: u64 = 0xa4093822299f31d0;
impl Default for FxHasher64 {
fn default() -> Self {
Self::new()
}
}
impl FxHasher64 {
pub const fn new() -> Self {
Self { hash: SEED }
}
#[inline]
fn add_to_hash(&mut self, i: u64) {
self.hash = self.hash.wrapping_add(i).wrapping_mul(K);
}
}
impl core::hash::Hasher for FxHasher64 {
#[inline]
fn write(&mut self, bytes: &[u8]) {
self.add_to_hash(hash_bytes(bytes));
}
#[inline]
fn write_u8(&mut self, i: u8) {
self.add_to_hash(i as u64);
}
#[inline]
fn write_u16(&mut self, i: u16) {
self.add_to_hash(i as u64);
}
#[inline]
fn write_u32(&mut self, i: u32) {
self.add_to_hash(i as u64);
}
#[inline]
fn write_u64(&mut self, i: u64) {
self.add_to_hash(i);
}
#[inline]
fn write_u128(&mut self, i: u128) {
self.add_to_hash(i as u64);
self.add_to_hash((i >> 64) as u64);
}
#[inline]
fn write_usize(&mut self, i: usize) {
self.add_to_hash(i as u64);
}
#[inline]
fn finish(&self) -> u64 {
self.hash.rotate_left(26)
}
}
#[inline]
fn multiply_mix(x: u64, y: u64) -> u64 {
let full = (x as u128) * (y as u128);
(full as u64) ^ ((full >> 64) as u64)
}
#[inline]
fn hash_bytes(bytes: &[u8]) -> u64 {
let len = bytes.len();
let mut s0 = SEED1;
let mut s1 = SEED2;
if len <= 16 {
if len >= 8 {
s0 ^= u64::from_le_bytes(bytes[0..8].try_into().unwrap());
s1 ^= u64::from_le_bytes(bytes[len - 8..].try_into().unwrap());
} else if len >= 4 {
s0 ^= u32::from_le_bytes(bytes[0..4].try_into().unwrap()) as u64;
s1 ^= u32::from_le_bytes(bytes[len - 4..].try_into().unwrap()) as u64;
} else if len > 0 {
let lo = bytes[0];
let mid = bytes[len / 2];
let hi = bytes[len - 1];
s0 ^= lo as u64;
s1 ^= ((hi as u64) << 8) | mid as u64;
}
} else {
let mut off = 0;
while off < len - 16 {
let x = u64::from_le_bytes(bytes[off..off + 8].try_into().unwrap());
let y = u64::from_le_bytes(bytes[off + 8..off + 16].try_into().unwrap());
let t = multiply_mix(s0 ^ x, PREVENT_TRIVIAL_ZERO_COLLAPSE ^ y);
s0 = s1;
s1 = t;
off += 16;
}
let suffix = &bytes[len - 16..];
s0 ^= u64::from_le_bytes(suffix[0..8].try_into().unwrap());
s1 ^= u64::from_le_bytes(suffix[8..16].try_into().unwrap());
}
multiply_mix(s0, s1) ^ (len as u64)
}
#[cfg(test)]
mod tests {
use super::FxHasher64;
use core::hash::Hasher;
#[test]
fn leading_zero_words_are_not_absorbed() {
let hash_words = |words: &[u64]| {
let mut hasher = FxHasher64::new();
for &word in words {
hasher.write_u64(word);
}
hasher.finish()
};
assert_ne!(hash_words(&[]), hash_words(&[0]), "a zero word must change the hash of the empty input");
assert_ne!(hash_words(&[0]), hash_words(&[0, 0]), "zero words must accumulate distinct states");
assert_ne!(hash_words(&[0, 7]), hash_words(&[7]), "a leading zero word must not be absorbed");
}
}