bevy_ecs/entity/
hash.rs

1use core::hash::{BuildHasher, Hasher};
2
3#[cfg(feature = "bevy_reflect")]
4use bevy_reflect::{std_traits::ReflectDefault, Reflect};
5
6/// A [`BuildHasher`] that results in a [`EntityHasher`].
7#[derive(Debug, Default, Clone)]
8#[cfg_attr(feature = "bevy_reflect", derive(Reflect), reflect(Default, Clone))]
9pub struct EntityHash;
10
11impl BuildHasher for EntityHash {
12    type Hasher = EntityHasher;
13
14    fn build_hasher(&self) -> Self::Hasher {
15        Self::Hasher::default()
16    }
17}
18
19/// A very fast hash that is only designed to work on generational indices
20/// like [`Entity`](super::Entity). It will panic if attempting to hash a type containing
21/// non-u64 fields.
22///
23/// This is heavily optimized for typical cases, where you have mostly live
24/// entities, and works particularly well for contiguous indices.
25///
26/// If you have an unusual case -- say all your indices are multiples of 256
27/// or most of the entities are dead generations -- then you might want also to
28/// try [`DefaultHasher`](bevy_platform::hash::DefaultHasher) for a slower hash
29/// computation but fewer lookup conflicts.
30#[derive(Debug, Default)]
31pub struct EntityHasher {
32    hash: u64,
33}
34
35impl Hasher for EntityHasher {
36    #[inline]
37    fn finish(&self) -> u64 {
38        self.hash
39    }
40
41    fn write(&mut self, _bytes: &[u8]) {
42        panic!("EntityHasher can only hash u64 fields.");
43    }
44
45    #[inline]
46    fn write_u64(&mut self, bits: u64) {
47        // SwissTable (and thus `hashbrown`) cares about two things from the hash:
48        // - H1: low bits (masked by `2ⁿ-1`) to pick the slot in which to store the item
49        // - H2: high 7 bits are used to SIMD optimize hash collision probing
50        // For more see <https://abseil.io/about/design/swisstables#metadata-layout>
51
52        // This hash function assumes that the entity ids are still well-distributed,
53        // so for H1 leaves the entity id alone in the low bits so that id locality
54        // will also give memory locality for things spawned together.
55        // For H2, take advantage of the fact that while multiplication doesn't
56        // spread entropy to the low bits, it's incredibly good at spreading it
57        // upward, which is exactly where we need it the most.
58
59        // While this does include the generation in the output, it doesn't do so
60        // *usefully*.  H1 won't care until you have over 3 billion entities in
61        // the table, and H2 won't care until something hits generation 33 million.
62        // Thus the comment suggesting that this is best for live entities,
63        // where there won't be generation conflicts where it would matter.
64
65        // The high 32 bits of this are ⅟φ for Fibonacci hashing.  That works
66        // particularly well for hashing for the same reason as described in
67        // <https://extremelearning.com.au/unreasonable-effectiveness-of-quasirandom-sequences/>
68        // It loses no information because it has a modular inverse.
69        // (Specifically, `0x144c_bc89_u32 * 0x9e37_79b9_u32 == 1`.)
70        //
71        // The low 32 bits make that part of the just product a pass-through.
72        const UPPER_PHI: u64 = 0x9e37_79b9_0000_0001;
73
74        // This is `(MAGIC * index + generation) << 32 + index`, in a single instruction.
75        self.hash = bits.wrapping_mul(UPPER_PHI);
76    }
77}
bevy_ecs/entity/hash.rs

bevy_ecs/entity/
hash.rs
