compiler/rustc_data_structures/src/fingerprint.rs

use std::hash::{Hash, Hasher};

use rustc_hashes::Hash64;
use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};

use crate::stable_hash::{FromStableHash, StableHasherHash, impl_stable_traits_for_trivial_type};

#[cfg(test)]
mod tests;

#[derive(Eq, PartialEq, Ord, PartialOrd, Debug, Clone, Copy)]
#[repr(C)]
pub struct Fingerprint(u64, u64);

pub trait FingerprintComponent {
    fn as_u64(&self) -> u64;
}

impl FingerprintComponent for Hash64 {
    #[inline]
    fn as_u64(&self) -> u64 {
        Hash64::as_u64(*self)
    }
}

impl FingerprintComponent for u64 {
    #[inline]
    fn as_u64(&self) -> u64 {
        *self
    }
}

impl Fingerprint {
    pub const ZERO: Fingerprint = Fingerprint(0, 0);

    #[inline]
    pub fn new<A, B>(_0: A, _1: B) -> Fingerprint
    where
        A: FingerprintComponent,
        B: FingerprintComponent,
    {
        Fingerprint(_0.as_u64(), _1.as_u64())
    }

    #[inline]
    pub fn to_smaller_hash(&self) -> Hash64 {
        // Even though both halves of the fingerprint are expected to be good
        // quality hash values, let's still combine the two values because the
        // Fingerprints in DefPathHash have the StableCrateId portion which is
        // the same for all DefPathHashes from the same crate. Combining the
        // two halves makes sure we get a good quality hash in such cases too.
        Hash64::new(self.0.wrapping_mul(3).wrapping_add(self.1))
    }

    #[inline]
    pub fn split(&self) -> (Hash64, Hash64) {
        (Hash64::new(self.0), Hash64::new(self.1))
    }

    #[inline]
    pub fn combine(self, other: Fingerprint) -> Fingerprint {
        // See https://stackoverflow.com/a/27952689 on why this function is
        // implemented this way.
        Fingerprint(
            self.0.wrapping_mul(3).wrapping_add(other.0),
            self.1.wrapping_mul(3).wrapping_add(other.1),
        )
    }

    #[inline]
    pub(crate) fn as_u128(self) -> u128 {
        u128::from(self.1) << 64 | u128::from(self.0)
    }

    // Combines two hashes in an order independent way. Make sure this is what
    // you want.
    #[inline]
    pub fn combine_commutative(self, other: Fingerprint) -> Fingerprint {
        let a = u128::from(self.1) << 64 | u128::from(self.0);
        let b = u128::from(other.1) << 64 | u128::from(other.0);

        let c = a.wrapping_add(b);

        Fingerprint(c as u64, (c >> 64) as u64)
    }

    pub fn to_hex(&self) -> String {
        format!("{:x}{:x}", self.0, self.1)
    }

    #[inline]
    pub fn to_le_bytes(&self) -> [u8; 16] {
        // This seems to optimize to the same machine code as
        // `unsafe { mem::transmute(*k) }`. Well done, LLVM! :)
        let mut result = [0u8; 16];

        let first_half: &mut [u8; 8] = (&mut result[0..8]).try_into().unwrap();
        *first_half = self.0.to_le_bytes();

        let second_half: &mut [u8; 8] = (&mut result[8..16]).try_into().unwrap();
        *second_half = self.1.to_le_bytes();

        result
    }

    #[inline]
    pub fn from_le_bytes(bytes: [u8; 16]) -> Fingerprint {
        Fingerprint(
            u64::from_le_bytes(bytes[0..8].try_into().unwrap()),
            u64::from_le_bytes(bytes[8..16].try_into().unwrap()),
        )
    }
}

impl std::fmt::Display for Fingerprint {
    fn fmt(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        write!(formatter, "{:x}-{:x}", self.0, self.1)
    }
}

impl Hash for Fingerprint {
    #[inline]
    fn hash<H: Hasher>(&self, state: &mut H) {
        state.write_fingerprint(self);
    }
}

trait FingerprintHasher {
    fn write_fingerprint(&mut self, fingerprint: &Fingerprint);
}

impl<H: Hasher> FingerprintHasher for H {
    #[inline]
    default fn write_fingerprint(&mut self, fingerprint: &Fingerprint) {
        self.write_u64(fingerprint.0);
        self.write_u64(fingerprint.1);
    }
}

impl FingerprintHasher for crate::unhash::Unhasher {
    #[inline]
    fn write_fingerprint(&mut self, fingerprint: &Fingerprint) {
        // Even though both halves of the fingerprint are expected to be good
        // quality hash values, let's still combine the two values because the
        // Fingerprints in DefPathHash have the StableCrateId portion which is
        // the same for all DefPathHashes from the same crate. Combining the
        // two halves makes sure we get a good quality hash in such cases too.
        //
        // Since `Unhasher` is used only in the context of HashMaps, it is OK
        // to combine the two components in an order-independent way (which is
        // cheaper than the more robust Fingerprint::to_smaller_hash()). For
        // HashMaps we don't really care if Fingerprint(x,y) and
        // Fingerprint(y, x) result in the same hash value. Collision
        // probability will still be much better than with FxHash.
        self.write_u64(fingerprint.0.wrapping_add(fingerprint.1));
    }
}

impl FromStableHash for Fingerprint {
    type Hash = StableHasherHash;

    #[inline]
    fn from(StableHasherHash([_0, _1]): Self::Hash) -> Self {
        Fingerprint(_0, _1)
    }
}

impl_stable_traits_for_trivial_type!(Fingerprint);

impl<E: Encoder> Encodable<E> for Fingerprint {
    #[inline]
    fn encode(&self, s: &mut E) {
        s.emit_raw_bytes(&self.to_le_bytes());
    }
}

impl<D: Decoder> Decodable<D> for Fingerprint {
    #[inline]
    fn decode(d: &mut D) -> Self {
        Fingerprint::from_le_bytes(d.read_raw_bytes(16).try_into().unwrap())
    }
}

/// `PackedFingerprint` wraps a `Fingerprint`.
/// Its purpose is to behave like a `Fingerprint` without alignment requirements.
///
/// This may be preferable to use in large collections of structs containing
/// fingerprints, as it can reduce memory consumption by preventing the padding
/// that the more strictly-aligned `Fingerprint` can introduce. An application of
/// this is in the query dependency graph, which contains a large collection of
/// `DepNode`s. As of this writing, the size of a `DepNode` decreases by 25%
/// (from 24 bytes to 18) by using the packed representation here, which
/// noticeably decreases total memory usage when compiling large crates.
///
/// (Unalignment was previously restricted to `x86` and `x86_64` hosts, but is
/// now enabled by default for all host architectures, in the hope that the
/// memory and cache savings should outweigh any unaligned access penalty.)
///
/// The wrapped `Fingerprint` is private to reduce the chance of a client
/// invoking undefined behavior by taking a reference to the packed field.
#[repr(packed)]
#[derive(Eq, PartialEq, Ord, PartialOrd, Debug, Clone, Copy, Hash)]
pub struct PackedFingerprint(Fingerprint);

impl std::fmt::Display for PackedFingerprint {
    #[inline]
    fn fmt(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        // Copy to avoid taking reference to packed field.
        let copy = self.0;
        copy.fmt(formatter)
    }
}

impl<E: Encoder> Encodable<E> for PackedFingerprint {
    #[inline]
    fn encode(&self, s: &mut E) {
        // Copy to avoid taking reference to packed field.
        let copy = self.0;
        copy.encode(s);
    }
}

impl<D: Decoder> Decodable<D> for PackedFingerprint {
    #[inline]
    fn decode(d: &mut D) -> Self {
        Self(Fingerprint::decode(d))
    }
}

impl From<Fingerprint> for PackedFingerprint {
    #[inline]
    fn from(f: Fingerprint) -> PackedFingerprint {
        PackedFingerprint(f)
    }
}

impl From<PackedFingerprint> for Fingerprint {
    #[inline]
    fn from(f: PackedFingerprint) -> Fingerprint {
        f.0
    }
}