Move hashes from rustc_data_structure to rustc_hashes so they can be shared with rust-analyzer

This commit is contained in:
Ben Kimock 2025-02-15 15:18:19 -05:00
parent 1d7cf0ff40
commit 4cf21866e8
50 changed files with 152 additions and 67 deletions

View file

@ -18,6 +18,7 @@ rustc-rayon = { version = "0.5.1", features = ["indexmap"] }
rustc-stable-hash = { version = "0.1.0", features = ["nightly"] }
rustc_arena = { path = "../rustc_arena" }
rustc_graphviz = { path = "../rustc_graphviz" }
rustc_hashes = { path = "../rustc_hashes" }
rustc_index = { path = "../rustc_index", package = "rustc_index" }
rustc_macros = { path = "../rustc_macros" }
rustc_serialize = { path = "../rustc_serialize" }

View file

@ -1,10 +1,9 @@
use std::hash::{Hash, Hasher};
use rustc_hashes::Hash64;
use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
use crate::stable_hasher::{
FromStableHash, Hash64, StableHasherHash, impl_stable_traits_for_trivial_type,
};
use crate::stable_hasher::{FromStableHash, StableHasherHash, impl_stable_traits_for_trivial_type};
#[cfg(test)]
mod tests;

View file

@ -1,152 +0,0 @@
//! rustc encodes a lot of hashes. If hashes are stored as `u64` or `u128`, a `derive(Encodable)`
//! will apply varint encoding to the hashes, which is less efficient than directly encoding the 8
//! or 16 bytes of the hash.
//!
//! The types in this module represent 64-bit or 128-bit hashes produced by a `StableHasher`.
//! `Hash64` and `Hash128` expose some utility functions to encourage users to not extract the inner
//! hash value as an integer type and accidentally apply varint encoding to it.
//!
//! In contrast with `Fingerprint`, users of these types cannot and should not attempt to construct
//! and decompose these types into constituent pieces. The point of these types is only to
//! connect the fact that they can only be produced by a `StableHasher` to their
//! `Encode`/`Decode` impls.
use std::fmt;
use std::ops::BitXorAssign;
use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
use crate::stable_hasher::{FromStableHash, StableHasherHash};
#[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Default)]
pub struct Hash64 {
inner: u64,
}
impl Hash64 {
pub const ZERO: Hash64 = Hash64 { inner: 0 };
#[inline]
pub fn new(n: u64) -> Self {
Self { inner: n }
}
#[inline]
pub fn as_u64(self) -> u64 {
self.inner
}
#[inline]
pub fn wrapping_add(self, other: Self) -> Self {
Self { inner: self.inner.wrapping_add(other.inner) }
}
}
impl BitXorAssign<u64> for Hash64 {
#[inline]
fn bitxor_assign(&mut self, rhs: u64) {
self.inner ^= rhs;
}
}
impl<S: Encoder> Encodable<S> for Hash64 {
#[inline]
fn encode(&self, s: &mut S) {
s.emit_raw_bytes(&self.inner.to_le_bytes());
}
}
impl<D: Decoder> Decodable<D> for Hash64 {
#[inline]
fn decode(d: &mut D) -> Self {
Self { inner: u64::from_le_bytes(d.read_raw_bytes(8).try_into().unwrap()) }
}
}
impl FromStableHash for Hash64 {
type Hash = StableHasherHash;
#[inline]
fn from(StableHasherHash([_0, __1]): Self::Hash) -> Self {
Self { inner: _0 }
}
}
impl fmt::Debug for Hash64 {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
self.inner.fmt(f)
}
}
impl fmt::LowerHex for Hash64 {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
fmt::LowerHex::fmt(&self.inner, f)
}
}
#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Default)]
pub struct Hash128 {
inner: u128,
}
// We expect Hash128 to be well mixed. So there's no point in hashing both parts.
//
// This also allows using Hash128-containing types in UnHash-based hashmaps, which would otherwise
// debug_assert! that we're hashing more than a single u64.
impl std::hash::Hash for Hash128 {
fn hash<H: std::hash::Hasher>(&self, h: &mut H) {
h.write_u64(self.truncate().as_u64());
}
}
impl Hash128 {
#[inline]
pub fn truncate(self) -> Hash64 {
Hash64 { inner: self.inner as u64 }
}
#[inline]
pub fn wrapping_add(self, other: Self) -> Self {
Self { inner: self.inner.wrapping_add(other.inner) }
}
#[inline]
pub fn as_u128(self) -> u128 {
self.inner
}
}
impl<S: Encoder> Encodable<S> for Hash128 {
#[inline]
fn encode(&self, s: &mut S) {
s.emit_raw_bytes(&self.inner.to_le_bytes());
}
}
impl<D: Decoder> Decodable<D> for Hash128 {
#[inline]
fn decode(d: &mut D) -> Self {
Self { inner: u128::from_le_bytes(d.read_raw_bytes(16).try_into().unwrap()) }
}
}
impl FromStableHash for Hash128 {
type Hash = StableHasherHash;
#[inline]
fn from(StableHasherHash([_0, _1]): Self::Hash) -> Self {
Self { inner: u128::from(_0) | (u128::from(_1) << 64) }
}
}
impl fmt::Debug for Hash128 {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
self.inner.fmt(f)
}
}
impl fmt::LowerHex for Hash128 {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
fmt::LowerHex::fmt(&self.inner, f)
}
}

View file

@ -84,7 +84,6 @@ pub mod vec_cache;
pub mod work_queue;
mod atomic_ref;
mod hashes;
/// This calls the passed function while ensuring it won't be inlined into the caller.
#[inline(never)]

View file

@ -10,12 +10,11 @@ use smallvec::SmallVec;
#[cfg(test)]
mod tests;
use rustc_hashes::{Hash64, Hash128};
pub use rustc_stable_hash::{
FromStableHash, SipHasher128Hash as StableHasherHash, StableSipHasher128 as StableHasher,
};
pub use crate::hashes::{Hash64, Hash128};
/// Something that implements `HashStable<CTX>` can be hashed in a way that is
/// stable across multiple compilation sessions.
///

View file

@ -1,7 +1,8 @@
use std::ptr;
use rustc_hashes::Hash128;
use super::*;
use crate::hashes::Hash128;
use crate::stable_hasher::{HashStable, StableHasher};
/// A tag type used in [`TaggedRef`] tests.