bitkit 3.2.2

High-performance, safe, no_std bit manipulation around a single Bits<T> newtype: checked APIs, BMI2-accelerated gather/scatter, Morton (Z-order) curves, zero runtime dependencies.
Documentation 
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

//! Medium-complexity workload benchmarks.
//!
//! Each benchmark exercises a realistic multi-step kernel rather than a
//! single arithmetic op:
//!
//! * `tsp15` — Held–Karp travelling-salesman DP on 15 cities, driven by
//!   the `submasks` iterator.
//! * `ipv4_parse_10k` — parse 10 000 IPv4 headers byte-for-byte.
//! * `sparse_extract` — pull 8 non-contiguous bit fields from a packed
//!   32-bit word.
//! * `popcount_sum_1M` — sum the population count of 1 million u64s.

#![allow(clippy::identity_op, clippy::needless_range_loop, clippy::erasing_op, clippy::unusual_byte_groupings)]

use criterion::{black_box, criterion_group, criterion_main, Criterion};

use bitkit::{bytes, Bits};

// =========================================================================
// 1. Held–Karp TSP DP over a 15-city distance matrix.
// =========================================================================
//
// dp[mask][last] = shortest path visiting the cities in `mask`, ending
// at `last`. Recurrence enumerates submasks of `mask` excluding `last`
// to find the predecessor. We compare two equivalent implementations:
//
// * `tsp_bitsrs` — uses `Bits::set_bits` to iterate predecessors and
//                   `Bits::has_bit` for set membership.
// * `tsp_inline` — uses bitwise ops directly.

const N: usize = 15;

fn random_distance_matrix(seed: u64) -> [[u32; N]; N] {
    let mut d = [[0u32; N]; N];
    let mut s = seed;
    for i in 0..N {
        for j in 0..N {
            if i == j { continue; }
            s = s.wrapping_mul(6364136223846793005).wrapping_add(1442695040888963407);
            d[i][j] = (s >> 33) as u32 % 1000 + 1;
        }
    }
    d
}

fn tsp_bitsrs(dist: &[[u32; N]; N]) -> u32 {
    let states = 1usize << N;
    let mut dp = vec![u32::MAX; states * N];
    dp[(1usize) * N + 0] = 0;

    for mask in 1u32..states as u32 {
        let mbits = Bits::<u32>::new(mask);
        if !mbits.has_bit(0).unwrap() { continue; }
        for last in mbits.set_bits() {
            if last == 0 && mask != 1 { continue; }
            let cur = dp[mask as usize * N + last as usize];
            if cur == u32::MAX { continue; }
            // Extend to every city `next` not yet in `mask`.
            let unvisited = !mask & ((1u32 << N) - 1);
            for next in Bits::<u32>::new(unvisited).set_bits() {
                let new_mask = mask | (1u32 << next);
                let cand = cur.saturating_add(dist[last as usize][next as usize]);
                let slot = &mut dp[new_mask as usize * N + next as usize];
                if cand < *slot { *slot = cand; }
            }
        }
    }
    let full = (1u32 << N) - 1;
    let mut best = u32::MAX;
    for last in 1..N {
        let val = dp[full as usize * N + last].saturating_add(dist[last][0]);
        if val < best { best = val; }
    }
    best
}

fn tsp_inline(dist: &[[u32; N]; N]) -> u32 {
    let states = 1usize << N;
    let mut dp = vec![u32::MAX; states * N];
    dp[1 * N + 0] = 0;
    for mask in 1u32..states as u32 {
        if mask & 1 == 0 { continue; }
        let mut m = mask;
        while m != 0 {
            let last = m.trailing_zeros();
            m &= m - 1;
            if last == 0 && mask != 1 { continue; }
            let cur = dp[mask as usize * N + last as usize];
            if cur == u32::MAX { continue; }
            let mut un = !mask & ((1u32 << N) - 1);
            while un != 0 {
                let next = un.trailing_zeros();
                un &= un - 1;
                let new_mask = mask | (1u32 << next);
                let cand = cur.saturating_add(dist[last as usize][next as usize]);
                let slot = &mut dp[new_mask as usize * N + next as usize];
                if cand < *slot { *slot = cand; }
            }
        }
    }
    let full = (1u32 << N) - 1;
    let mut best = u32::MAX;
    for last in 1..N {
        let val = dp[full as usize * N + last].saturating_add(dist[last][0]);
        if val < best { best = val; }
    }
    best
}

fn tsp_bitsrs_foreach(dist: &[[u32; N]; N]) -> u32 {
    let states = 1usize << N;
    let mut dp = vec![u32::MAX; states * N];
    dp[(1usize) * N + 0] = 0;
    for mask in 1u32..states as u32 {
        let mbits = Bits::<u32>::new(mask);
        if !mbits.has_bit(0).unwrap() { continue; }
        mbits.set_bits().for_each(|last| {
            if last == 0 && mask != 1 { return; }
            let cur = dp[mask as usize * N + last as usize];
            if cur == u32::MAX { return; }
            let unvisited = !mask & ((1u32 << N) - 1);
            Bits::<u32>::new(unvisited).set_bits().for_each(|next| {
                let new_mask = mask | (1u32 << next);
                let cand = cur.saturating_add(dist[last as usize][next as usize]);
                let slot = &mut dp[new_mask as usize * N + next as usize];
                if cand < *slot { *slot = cand; }
            });
        });
    }
    let full = (1u32 << N) - 1;
    let mut best = u32::MAX;
    for last in 1..N {
        let val = dp[full as usize * N + last].saturating_add(dist[last][0]);
        if val < best { best = val; }
    }
    best
}

fn bench_tsp(c: &mut Criterion) {
    let d = random_distance_matrix(0xDEAD_BEEF_CAFE_BABE);
    let mut grp = c.benchmark_group("tsp15");
    grp.sample_size(20); // each iteration is heavy
    grp.bench_function("bits-rs (submasks/set_bits)", |b| {
        b.iter(|| black_box(tsp_bitsrs(black_box(&d))))
    });
    grp.bench_function("inline (raw ops)", |b| {
        b.iter(|| black_box(tsp_inline(black_box(&d))))
    });
    grp.bench_function("bits-rs (for_each fast path)", |b| {
        b.iter(|| black_box(tsp_bitsrs_foreach(black_box(&d))))
    });
    grp.finish();
}

// =========================================================================
// 2. Parse 10 000 IPv4 headers.
// =========================================================================

const PACKETS: usize = 10_000;

fn make_packets() -> Vec<[u8; 20]> {
    (0..PACKETS).map(|i| {
        let i32_ = i as u32;
        [
            0x45, 0x00, (i32_ >> 8) as u8, i32_ as u8,
            (i32_ >> 0) as u8, (i32_ >> 8) as u8, 0x40, 0x00,
            64, 6, 0xAB, 0xCD,
            192, 168, (i >> 8) as u8, i as u8,
            8, 8, 8, 8,
        ]
    }).collect()
}

#[derive(Default)]
struct ParsedSummary {
    version_sum: u64,
    ihl_sum: u64,
    total_length_sum: u64,
    ttl_sum: u64,
    src_xor: u32,
}

fn parse_bitsrs(pkts: &[[u8; 20]]) -> ParsedSummary {
    let mut s = ParsedSummary::default();
    for hdr in pkts {
        let b0 = Bits::<u8>::new(hdr[0]);
        s.version_sum += b0.extract(4..8).unwrap().get() as u64;
        s.ihl_sum     += b0.extract(0..4).unwrap().get() as u64;
        s.total_length_sum += bytes::read_u16_be(&hdr[2..]).unwrap() as u64;
        s.ttl_sum     += hdr[8] as u64;
        s.src_xor     ^= bytes::read_u32_be(&hdr[12..]).unwrap();
    }
    s
}

fn parse_inline(pkts: &[[u8; 20]]) -> ParsedSummary {
    let mut s = ParsedSummary::default();
    for hdr in pkts {
        s.version_sum += (hdr[0] >> 4) as u64;
        s.ihl_sum     += (hdr[0] & 0xF) as u64;
        s.total_length_sum += u16::from_be_bytes([hdr[2], hdr[3]]) as u64;
        s.ttl_sum     += hdr[8] as u64;
        s.src_xor     ^= u32::from_be_bytes([hdr[12], hdr[13], hdr[14], hdr[15]]);
    }
    s
}

fn bench_ipv4(c: &mut Criterion) {
    let pkts = make_packets();
    let mut grp = c.benchmark_group("ipv4_parse_10k");
    grp.bench_function("bits-rs (bytes::read + Bits::extract)", |b| {
        b.iter(|| black_box(parse_bitsrs(black_box(&pkts))));
    });
    grp.bench_function("inline (shift/and/from_be_bytes)", |b| {
        b.iter(|| black_box(parse_inline(black_box(&pkts))));
    });
    grp.finish();
}

// =========================================================================
// 3. Sparse bit extract — pull 8 non-contiguous fields from a 32-bit word
//    using one gather (PEXT) vs a shift+and chain.
// =========================================================================
//
// Layout of the 32-bit "message":
//   bit 31:    flag A
//   bits 30-24: opcode (7 bits)
//   bit 23:    flag B
//   bits 22-16: tag (7 bits)
//   bit 15:    flag C
//   bits 14-8: payload-hi (7 bits)
//   bit 7:     flag D
//   bits 6-0:  payload-lo (7 bits)
//
// Gather mask 0xFFFF_FFFF returns the raw 32 bits packed (useless), but
// a realistic example wants the 4 flags packed into a nibble. That's
// what `gather(0x8080_8080)` does in one instruction on BMI2.

fn extract_flags_bitsrs(xs: &[u32]) -> u32 {
    let mask = Bits::<u32>::new(0x8080_8080);
    let mut acc = 0u32;
    for &x in xs {
        acc ^= Bits::<u32>::new(x).gather(mask).get();
    }
    acc
}

fn extract_flags_inline(xs: &[u32]) -> u32 {
    let mut acc = 0u32;
    for &x in xs {
        let a = (x >> 31) & 1;
        let b = (x >> 23) & 1;
        let c = (x >> 15) & 1;
        let d = (x >> 7) & 1;
        acc ^= a | (b << 1) | (c << 2) | (d << 3);
    }
    acc
}

fn bench_sparse_extract(c: &mut Criterion) {
    let xs: Vec<u32> = (0..10_000u32).map(|i| i.wrapping_mul(0x9E37_79B1)).collect();
    let mut grp = c.benchmark_group("sparse_extract_10k");
    grp.bench_function("bits-rs (single gather/PEXT)", |b| {
        b.iter(|| black_box(extract_flags_bitsrs(black_box(&xs))))
    });
    grp.bench_function("inline (shift+and×4)", |b| {
        b.iter(|| black_box(extract_flags_inline(black_box(&xs))))
    });
    grp.finish();
}

// =========================================================================
// 4. Popcount sum over 1 M u64s.
// =========================================================================

fn popcount_sum_bitsrs(xs: &[u64]) -> u64 {
    let mut s: u64 = 0;
    for &x in xs { s = s.wrapping_add(Bits::<u64>::new(x).count_ones() as u64); }
    s
}
fn popcount_sum_inline(xs: &[u64]) -> u64 {
    let mut s: u64 = 0;
    for &x in xs { s = s.wrapping_add(x.count_ones() as u64); }
    s
}

fn bench_popcount_sum(c: &mut Criterion) {
    let xs: Vec<u64> = (0..1_000_000u64)
        .map(|i| i.wrapping_mul(0x9E37_79B97F4A_7C15))
        .collect();
    let mut grp = c.benchmark_group("popcount_sum_1M_u64");
    grp.bench_function("bits-rs", |b| b.iter(|| black_box(popcount_sum_bitsrs(black_box(&xs)))));
    grp.bench_function("primitive", |b| b.iter(|| black_box(popcount_sum_inline(black_box(&xs)))));
    grp.finish();
}

criterion_group!(workloads, bench_tsp, bench_ipv4, bench_sparse_extract, bench_popcount_sum);
criterion_main!(workloads);