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

use aligned_vec::typenum::U32;
use aligned_vec::AlignedVec;
use bitpolymul_sys::{
    bc_to_lch_2_unit256, bc_to_mono_2_unit256, btfy_128, decode_128, encode_128_half_input_zero,
    gf2ext128_mul_sse, i_btfy_128,
};
use std::cmp::max;

#[derive(Clone, Debug, Default)]
pub struct FftPoly {
    n: usize,
    n_pow2: usize,
    poly: AlignedVec<u64, U32>,
}

#[derive(Default, Clone)]
pub struct DecodeCache {
    temp: AlignedVec<u64, U32>,
}

impl FftPoly {
    pub fn new() -> Self {
        Self::default()
    }

    pub fn resize(&mut self, n: usize) {
        if self.n == n {
            // do nothing
        } else if n == 0 {
            self.n = 0;
            self.n_pow2 = 0;
            self.poly.clear();
        } else {
            self.n = n;
            let log_n = log2_ceil(self.n);
            self.n_pow2 = max(1 << log_n, 256);
            self.poly.resize(2 * self.n_pow2, 0);
        }
    }

    pub fn encode(&mut self, data: &[u64]) {
        self.resize(data.len());
        if self.n == 0 {
            return;
        }
        let log_n = log2_ceil(self.n_pow2);
        let mut temp: AlignedVec<_, U32> = AlignedVec::from(data);
        temp.resize(self.n_pow2, 0);
        let n_pow2 = self.n_pow2.try_into().expect("n_pow2 overflows u32");
        unsafe {
            bc_to_lch_2_unit256(temp.as_mut_ptr(), n_pow2);
            encode_128_half_input_zero(self.poly.as_mut_ptr(), temp.as_ptr(), n_pow2);
            btfy_128(self.poly.as_mut_ptr(), n_pow2, 64 + log_n + 1);
        }
    }

    pub fn encode_new(data: &[u64]) -> Self {
        let mut fft = Self::new();
        fft.encode(data);
        fft
    }

    pub fn mult_eq(&mut self, a: &FftPoly) {
        self.mult(&self.clone(), a)
    }

    pub fn mult(&mut self, a: &FftPoly, b: &FftPoly) {
        assert_eq!(a.n_pow2, b.n_pow2);
        self.resize(a.n);
        unsafe {
            for i in 0..self.n_pow2 {
                gf2ext128_mul_sse(
                    &mut self.poly[i * 2] as *mut _ as *mut _,
                    &a.poly[i * 2] as *const _ as *const _,
                    &b.poly[i * 2] as *const _ as *const _,
                );
            }
        }
    }
    pub fn add_eq(&mut self, a: &FftPoly) {
        self.add(&self.clone(), a)
    }

    pub fn add(&mut self, a: &FftPoly, b: &FftPoly) {
        assert_eq!(a.n_pow2, b.n_pow2);
        self.resize(a.n);
        self.poly
            .iter_mut()
            .zip(a.poly.iter())
            .zip(b.poly.iter())
            .for_each(|((this, a), b)| *this = *a ^ *b);
    }

    pub fn decode(self, dest: &mut [u64]) {
        let mut cache = DecodeCache::default();
        self.decode_with_cache(&mut cache, dest)
    }

    pub fn decode_with_cache(mut self, cache: &mut DecodeCache, dest: &mut [u64]) {
        assert_eq!(dest.len(), 2 * self.n);
        if cache.temp.len() < self.poly.len() {
            cache.temp.resize(self.poly.len(), 0);
        }
        let log_n = log2_ceil(self.n_pow2);
        let n_pow2 = self.n_pow2.try_into().expect("n_pow2 overflow");
        unsafe {
            i_btfy_128(self.poly.as_mut_ptr(), n_pow2, 64 + log_n + 1);
            decode_128(cache.temp.as_mut_ptr(), self.poly.as_ptr(), n_pow2);
            bc_to_mono_2_unit256(cache.temp.as_mut_ptr(), 2 * n_pow2);
        }
        dest.copy_from_slice(&cache.temp[..dest.len()])
    }
}

pub(crate) fn log2_floor(val: usize) -> u32 {
    assert!(val > 0);
    usize::BITS - val.leading_zeros() - 1
}

pub(crate) fn log2_ceil(val: usize) -> u32 {
    let floor = log2_floor(val);
    if val > (1 << floor) {
        floor + 1
    } else {
        floor
    }
}

#[cfg(test)]
mod tests {
    use crate::FftPoly;
    use rand::{thread_rng, Rng};

    pub(crate) fn bitpolymul(c: &mut [u64], a: &[u64], b: &[u64]) {
        assert_eq!(a.len(), b.len());
        let mut a = FftPoly::encode_new(a);
        let b = FftPoly::encode_new(b);
        debug_assert_eq!(a.poly.as_ptr() as usize % 32, 0);
        debug_assert_eq!(b.poly.as_ptr() as usize % 32, 0);
        a.mult_eq(&b);
        a.decode(c)
    }

    #[test]
    fn basic_test() {
        let len = 1 << 12;
        let mut poly1 = vec![0_u64; len];
        let mut poly2 = vec![0_u64; len];
        let mut rpoly1 = vec![0_u64; 2 * len];
        let mut rpoly2 = vec![0_u64; 2 * len];
        thread_rng().fill(&mut poly1[..]);
        thread_rng().fill(&mut poly2[..]);
        bitpolymul(&mut rpoly1, &poly1, &poly2);
        bitpolymul(&mut rpoly2, &poly2, &poly1);
        assert_eq!(rpoly1, rpoly2);
    }
}