1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
use std::{error, fmt, mem, slice};

fn naive(x: &[u8], y: &[u8]) -> u64 {
    assert_eq!(x.len(), y.len());
    x.iter().zip(y).fold(0, |a, (b, c)| a + (*b ^ *c).count_ones() as u64)
}

#[derive(PartialEq, Eq, Ord, PartialOrd, Hash, Clone)]
pub struct DistanceError {
    _x: ()
}
impl fmt::Debug for DistanceError {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        write!(f, "hamming::distance_fast: called with slices with different 8-byte alignments")
    }
}
impl fmt::Display for DistanceError {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        fmt::Debug::fmt(self, f)
    }
}
impl error::Error for DistanceError {
    fn description(&self) -> &str {
        "hamming::distance_fast: called with slices with different 8-byte alignments"
    }
}

/// Computes the bitwise [Hamming
/// distance](https://en.wikipedia.org/wiki/Hamming_distance) between
/// `x` and `y`, that is, the number of bits where `x` and `y` differ,
/// or, the number of set bits in the xor of `x` and `y`.
///
/// This is a highly optimised version of the following naive version:
///
/// ```rust
/// fn naive(x: &[u8], y: &[u8]) -> u64 {
///     x.iter().zip(y).fold(0, |a, (b, c)| a + (*b ^ *c).count_ones() as u64)
/// }
/// ```
///
/// This function requires that `x` and `y` have the same 8-byte
/// alignment (that is, `x.as_ptr() as usize % 8 == y.as_ptr() as
/// usize % 8`). If not, `Err` is returned. If sub-optimal performance
/// can be tolerated, consider using `distance` which incorporates a
/// fallback to a slower but less restrictive algorithm.
///
/// It is essentially guaranteed that `x` and `y` will have the same
/// 8-byte alignment if they are both just `Vec<u8>`s of non-trivial
/// length (e.g. larger than 8) as in the example below.
///
/// This is implemented using the same tree-merging approach as
/// `weight`, see there for details.
///
/// # Panics
///
/// `x` and `y` must have the same length, or else `distance_fast` panics.
///
/// # Performance Comparison
///
/// | length | `naive` (ns) | `distance_fast` (ns) | `naive`/`distance_fast` |
/// |--:|--:|--:|--:|
/// | 1 | 5  | 6  | 0.83 |
/// | 10 | 44  | 45  | 0.97 |
/// | 100 | 461  | 473  | 0.97 |
/// | 1,000 | 4,510  | 397  | 11 |
/// | 10,000 | 46,700  | 2,740  | 17 |
/// | 100,000 | 45,600  | 20,400  | 22 |
/// | 1,000,000 | 4,590,000  | 196,000  | 23 |
///
/// # Examples
///
/// ```rust
/// let x = vec![0xFF; 1000];
/// let y = vec![0; 1000];
/// assert_eq!(hamming::distance_fast(&x, &y), Ok(8 * 1000));
///
/// // same alignment, but moderately complicated
/// assert_eq!(hamming::distance_fast(&x[1..1000 - 8], &y[8 + 1..]), Ok(8 * (1000 - 8 - 1)));
///
/// // differing alignments
/// assert!(hamming::distance_fast(&x[1..], &y[..999]).is_err());
/// ```
pub fn distance_fast(x: &[u8], y: &[u8]) -> Result<u64, DistanceError> {
    assert_eq!(x.len(), y.len());

    const M1: u64 = 0x5555555555555555;
    const M2: u64 = 0x3333333333333333;
    const M4: u64 = 0x0F0F0F0F0F0F0F0F;
    const M8: u64 = 0x00FF00FF00FF00FF;

    type T30 = [u64; 30];
    let size = mem::size_of::<T30>();
    let alignment = mem::align_of::<T30>();

    let ptr1 = x.as_ptr() as usize;
    let ptr2 = y.as_ptr() as usize;

    // round up to the nearest multiple
    let aligned1 = (ptr1 + alignment - 1) / alignment * alignment;
    let aligned2 = (ptr2 + alignment - 1) / alignment * alignment;
    let distance1 = aligned1 - ptr1;
    let distance2 = aligned2 - ptr2;

    // can't fit a single T30 in
    if x.len() < size + distance1 {
        return Ok(naive(x, y))
    }

    if distance1 != distance2 {
        return Err(DistanceError { _x: () });
    }

    let (head1, middle1) = x.split_at(distance1);
    let (head2, middle2) = y.split_at(distance2);

    assert!(middle1.as_ptr() as usize % alignment == 0);
    assert!(middle2.as_ptr() as usize % alignment == 0);
    let (thirty1, thirty2) = unsafe {
        (slice::from_raw_parts(middle1.as_ptr() as *const T30,
                               middle1.len() / size),
         slice::from_raw_parts(middle2.as_ptr() as *const T30,
                               middle2.len() / size))
    };
    debug_assert_eq!(thirty1.len(), thirty2.len());

    let tail1 = &middle1[thirty1.len() * size..];
    let tail2 = &middle2[thirty2.len() * size..];
    let mut count = naive(head1, head2) + naive(tail1, tail2);
    for (array1, array2) in thirty1.iter().zip(thirty2) {
        let mut acc = 0;
        for j_ in 0..10 {
            let j = j_ * 3;
            let mut count1 = array1[j] ^ array2[j];
            let mut count2 = array1[j + 1] ^ array2[j + 1];
            let mut half1 = array1[j + 2] ^ array2[j + 2];
            let mut half2 = half1;
            half1 &= M1;
            half2 = (half2 >> 1) & M1;
            count1 -= (count1 >> 1) & M1;
            count2 -= (count2 >> 1) & M1;
            count1 += half1;
            count2 += half2;
            count1 = (count1 & M2) + ((count1 >> 2) & M2);
            count1 += (count2 & M2) + ((count2 >> 2) & M2);
            acc += (count1 & M4) + ((count1 >> 4) & M4);
        }
        acc = (acc & M8) + ((acc >> 8) & M8);
        acc =  acc       +  (acc >> 16);
        acc =  acc       +  (acc >> 32);
        count += acc & 0xFFFF;
    }
    Ok(count)
}

/// Computes the bitwise [Hamming
/// distance](https://en.wikipedia.org/wiki/Hamming_distance) between
/// `x` and `y`, that is, the number of bits where `x` and `y` differ,
/// or, the number of set bits in the xor of `x` and `y`.
///
/// When `x` and `y` have the same 8-byte alignment, this uses
/// `distance_fast`, a highly optimised version of the following naive
/// version:
///
/// ```rust
/// fn naive(x: &[u8], y: &[u8]) -> u64 {
///     x.iter().zip(y).fold(0, |a, (b, c)| a + (*b ^ *c).count_ones() as u64)
/// }
/// ```
///
/// If alignments differ, a slower but less restrictive algorithm is
/// used.
///
/// It is essentially guaranteed that `x` and `y` will have the same
/// 8-byte alignment if they are both just `Vec<u8>`s of non-trivial
/// length (e.g. larger than 8) as in the example below.
///
/// # Panics
///
/// `x` and `y` must have the same length, or else `distance` panics.
///
/// # Performance Comparison
///
/// | length | `naive` (ns) | `distance` (ns) | `naive`/`distance` |
/// |--:|--:|--:|--:|
/// | 1 | 5  | 6  | 0.83 |
/// | 10 | 44  | 45  | 0.97 |
/// | 100 | 461  | 473  | 0.97 |
/// | 1,000 | 4,510  | 397  | 11 |
/// | 10,000 | 46,700  | 2,740  | 17 |
/// | 100,000 | 45,600  | 20,400  | 22 |
/// | 1,000,000 | 4,590,000  | 196,000  | 23 |
///
/// The benchmarks ensured that `x` and `y` had the same alignment.
///
/// # Examples
///
/// ```rust
/// let x = vec![0xFF; 1000];
/// let y = vec![0; 1000];
/// assert_eq!(hamming::distance(&x, &y), 8 * 1000);
/// ```
pub fn distance(x: &[u8], y: &[u8]) -> u64 {
    distance_fast(x, y)
        .ok()
        .unwrap_or_else(|| naive(x, y))
}

#[cfg(all(test, feature = "unstable"))]
fn distance_(x: &[u8], y: &[u8]) -> u64 {
    distance_fast(x, y).unwrap()
}

#[cfg(test)]
mod tests {
    use quickcheck as qc;
    use rand;
    #[test]
    fn naive_smoke() {
        let tests: &[(&[u8], &[u8], u64)] = &[
            (&[], &[], 0),
            (&[0], &[0], 0),
            (&[0], &[0xFF], 8),
            (&[0b10101010], &[0b01010101], 8),
            (&[0b11111010], &[0b11110101], 4),
            (&[0; 10], &[0; 10], 0),
            (&[0xFF; 10], &[0x0F; 10], 4 * 10),
            (&[0x3B; 10000], &[0x3B; 10000], 0),
            (&[0x77; 10000], &[0x3B; 10000], 3 * 10000),
            ];
        for &(x, y, expected) in tests {
            assert_eq!(super::naive(x, y), expected);
        }
    }
    #[test]
    fn distance_fast_qc() {
        fn prop(v: Vec<u8>, w: Vec<u8>, misalign: u8) -> qc::TestResult {
            let l = ::std::cmp::min(v.len(), w.len());
            if l < misalign as usize {
                return qc::TestResult::discard()
            }

            let x = &v[misalign as usize..l];
            let y = &w[misalign as usize..l];
            qc::TestResult::from_bool(super::distance_fast(x, y).unwrap() == super::naive(x, y))
        }
        qc::QuickCheck::new()
            .gen(qc::StdGen::new(rand::thread_rng(), 10_000))
            .quickcheck(prop as fn(Vec<u8>,Vec<u8>,u8) -> qc::TestResult)
    }
    #[test]
    fn distance_fast_smoke_huge() {
        let v = vec![0b1001_1101; 10234567];
        let w = vec![0b1111_1111; v.len()];

        assert_eq!(super::distance_fast(&v, &v).unwrap(), 0);
        assert_eq!(super::distance_fast(&v, &w).unwrap(), 3 * w.len() as u64);
    }
    #[test]
    fn distance_smoke() {
        let v = vec![0; 10000];
        let w = vec![0xFF; v.len()];
        for len_ in 0..99 {
            let len = len_ * 10;
            for i in 0..8 {
                for j in 0..8 {
                    assert_eq!(super::distance(&v[i..i+len], &w[j..j+len]),
                               len as u64 * 8)
                }
            }
        }
    }
}

#[cfg(all(test, feature = "unstable"))]
mod benches {
    use test;
    fn bench<F: FnMut(&[u8], &[u8]) -> u64>(b: &mut test::Bencher, n: usize, mut f: F) {
        let data = vec![0xFF; n];
        b.iter(|| {
            let d1 = test::black_box(&data);
            let d2 = test::black_box(&data);
            f(d1, d2)
        })
    }
    macro_rules! test_mod {
        ($name: ident) => {
            mod $name {
                use test;
                use super::bench;
                use super::super::$name;
                #[bench]
                fn _0000001(b: &mut test::Bencher) { bench(b, 1, $name) }
                #[bench]
                fn _0000010(b: &mut test::Bencher) { bench(b, 10, $name) }
                #[bench]
                fn _0000100(b: &mut test::Bencher) { bench(b, 100, $name) }
                #[bench]
                fn _0001000(b: &mut test::Bencher) { bench(b, 1000, $name) }
                #[bench]
                fn _0010000(b: &mut test::Bencher) { bench(b, 10000, $name) }
                #[bench]
                fn _0100000(b: &mut test::Bencher) { bench(b, 100000, $name) }
                #[bench]
                fn _1000000(b: &mut test::Bencher) { bench(b, 1000000, $name) }
            }
        }
    }
    test_mod!(naive);
    test_mod!(distance_);
}