aoc/year2015/
day04.rs

1//! # The Ideal Stocking Stuffer
2//!
3//! This solution relies on brute forcing combinations as quickly as possible using an internal
4//! implementation of the [`MD5`] hashing algorithm.
5//!
6//! Each number's hash is independent of the others, so we speed things up by using threading
7//! to search in parallel in blocks of 1000 numbers at a time.
8//!
9//! Using the [`format!`] macro to join the secret key to the number is quite slow. To go faster
10//! we reuse the same `u8` buffer, incrementing digits one at a time.
11//! The numbers from 1 to 999 are handled specially.
12//!
13//! Interestingly, the total time to solve this problem is *extremely* sensitive to the secret key
14//! provided as input. For example, my key required ~10⁷ iterations to find the answer to part two.
15//! However, for unit testing, I was able to randomly find a value that takes only 455 iterations,
16//! about 22,000 times faster!
17//!
18//! [`MD5`]: crate::util::md5
19//! [`format!`]: std::format
20use crate::util::md5::*;
21use crate::util::thread::*;
22use implementation::*;
23use std::sync::atomic::{AtomicU32, Ordering};
24
25pub struct Shared {
26    prefix: String,
27    iter: AtomicIter,
28    first: AtomicU32,
29    second: AtomicU32,
30}
31
32pub fn parse(input: &str) -> Shared {
33    let shared = Shared {
34        prefix: input.trim().to_owned(),
35        iter: AtomicIter::new(1000, 1000),
36        first: AtomicU32::new(u32::MAX),
37        second: AtomicU32::new(u32::MAX),
38    };
39
40    // Handle the first 999 numbers specially as the number of digits varies.
41    for n in 1..1000 {
42        let (mut buffer, size) = format_string(&shared.prefix, n);
43        check_hash(&mut buffer, size, n, &shared);
44    }
45
46    // Use as many cores as possible to parallelize the remaining search.
47    spawn(|| worker(&shared));
48    shared
49}
50
51pub fn part1(input: &Shared) -> u32 {
52    input.first.load(Ordering::Relaxed)
53}
54
55pub fn part2(input: &Shared) -> u32 {
56    input.second.load(Ordering::Relaxed)
57}
58
59fn format_string(prefix: &str, mut n: u32) -> ([u8; 64], usize) {
60    let prefix_len = prefix.len();
61    let digits = n.max(1).ilog10() as usize + 1;
62    let size = prefix_len + digits;
63
64    let mut buffer = [0; 64];
65    buffer[..prefix_len].copy_from_slice(prefix.as_bytes());
66
67    for i in (prefix_len..size).rev() {
68        buffer[i] = b'0' + (n % 10) as u8;
69        n /= 10;
70    }
71
72    (buffer, size)
73}
74
75fn check_hash(buffer: &mut [u8], size: usize, n: u32, shared: &Shared) {
76    let [result, ..] = hash(buffer, size);
77
78    if result & 0xffffff00 == 0 {
79        shared.second.fetch_min(n, Ordering::Relaxed);
80        shared.iter.stop();
81    } else if result & 0xfffff000 == 0 {
82        shared.first.fetch_min(n, Ordering::Relaxed);
83    }
84}
85
86#[cfg(not(feature = "simd"))]
87mod implementation {
88    use super::*;
89
90    pub(super) fn worker(shared: &Shared) {
91        while let Some(offset) = shared.iter.next() {
92            let (mut buffer, size) = format_string(&shared.prefix, offset);
93
94            for n in 0..1000 {
95                // Format macro is very slow, so update digits directly.
96                buffer[size - 3] = b'0' + (n / 100) as u8;
97                buffer[size - 2] = b'0' + ((n / 10) % 10) as u8;
98                buffer[size - 1] = b'0' + (n % 10) as u8;
99
100                check_hash(&mut buffer, size, offset + n, shared);
101            }
102        }
103    }
104}
105
106#[cfg(feature = "simd")]
107mod implementation {
108    use super::*;
109    use crate::util::bitset::*;
110    use crate::util::md5::simd::hash_fixed;
111    use std::simd::cmp::SimdPartialEq as _;
112    use std::simd::*;
113
114    #[expect(clippy::needless_range_loop)]
115    fn check_hash_simd<const N: usize>(
116        buffers: &mut [[u8; 64]; N],
117        size: usize,
118        start: u32,
119        offset: u32,
120        shared: &Shared,
121    ) {
122        // Format macro is very slow, so update digits directly.
123        for i in 0..N {
124            let n = offset + i as u32;
125            buffers[i][size - 3] = b'0' + (n / 100) as u8;
126            buffers[i][size - 2] = b'0' + ((n / 10) % 10) as u8;
127            buffers[i][size - 1] = b'0' + (n % 10) as u8;
128        }
129
130        let [result, ..] = hash_fixed(buffers, size);
131        let bitmask = (result & Simd::splat(0xfffff000)).simd_eq(Simd::splat(0)).to_bitmask();
132
133        for i in bitmask.biterator() {
134            if result[i] & 0xffffff00 == 0 {
135                shared.second.fetch_min(start + offset + i as u32, Ordering::Relaxed);
136                shared.iter.stop();
137            } else {
138                shared.first.fetch_min(start + offset + i as u32, Ordering::Relaxed);
139            }
140        }
141    }
142
143    pub(super) fn worker(shared: &Shared) {
144        while let Some(start) = shared.iter.next() {
145            let (prefix, size) = format_string(&shared.prefix, start);
146            let buffers = &mut [prefix; 32];
147
148            for offset in (0..992).step_by(32) {
149                check_hash_simd(buffers, size, start, offset, shared);
150            }
151
152            let buffers = &mut [prefix; 8];
153            check_hash_simd(buffers, size, start, 992, shared);
154        }
155    }
156}
aoc/year2015/day04.rs

aoc/year2015/
day04.rs
