urng 0.4.6

Universal Random Number Generator
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

use crate::{_internal::FSCALE32, rng::Rng32, rng32::SplitMix32};

#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;

/// JSF (Jenkins Small Fast) 32-bit RNG implementation.
///
/// # Example
/// ```
/// use urng::rng::Rng32;
/// use urng::rng32::Jsf32;
///
/// let mut rng = Jsf32::new(12345);
/// ```
#[repr(C, align(64))]
pub struct Jsf32 {
    pub(crate) a: u32,
    pub(crate) b: u32,
    pub(crate) c: u32,
    pub(crate) d: u32,
}

impl Jsf32 {
    /// Creates a new `Jsf32` instance with the given seed.
    pub fn new(seed: u32) -> Self {
        let mut seedgen = SplitMix32::new(seed);
        let seed = seedgen.nextu();
        Self {
            a: 0xf1ea5eed,
            b: seed,
            c: seed,
            d: seed,
        }
    }
}

impl Rng32 for Jsf32 {
    #[inline(always)]
    fn nextu(&mut self) -> u32 {
        let e = self.a.wrapping_sub(self.b.rotate_left(27));
        self.a = self.b ^ self.c.rotate_left(17);
        self.b = self.c.wrapping_add(self.d);
        self.c = self.d.wrapping_add(e);
        self.d = e.wrapping_add(self.a);
        self.d
    }
}

/// 16-way SIMD implementation of JSF (Jenkins Small Fast) 32-bit RNG.
/// This implementation uses AVX-512 instructions to generate 16 random numbers in parallel.
///
/// # Example
/// ```
/// use urng::rng::Rng32;
/// use urng::rng32::Jsf32x16;
///
/// let mut rng = unsafe { Jsf32x16::new(12345) };
/// let _ = unsafe { rng.nextu() };
/// ```
#[repr(C, align(64))]
pub struct Jsf32x16 {
    pub(crate) a: __m512i,
    pub(crate) b: __m512i,
    pub(crate) c: __m512i,
    pub(crate) d: __m512i,
}

pub(crate) const JSF32X16: usize = 16;

impl Jsf32x16 {
    #[target_feature(enable = "avx512f")]
    pub unsafe fn new(seed: u32) -> Self {
        let mut seedgen = SplitMix32::new(seed);
        let mut sv = [[0u32; JSF32X16]; 3];
        for vals in sv.iter_mut() {
            for v in vals.iter_mut() {
                *v = seedgen.nextu();
            }
        }
        let a = [0xf1ea5eedu32; JSF32X16];
        unsafe {
            Self {
                a: _mm512_loadu_si512(a.as_ptr() as *const __m512i),
                b: _mm512_loadu_si512(sv[0].as_ptr() as *const __m512i),
                c: _mm512_loadu_si512(sv[1].as_ptr() as *const __m512i),
                d: _mm512_loadu_si512(sv[2].as_ptr() as *const __m512i),
            }
        }
    }

    #[inline]
    #[target_feature(enable = "avx512f")]
    pub unsafe fn nextu_vec(&mut self) -> __m512i {
        let e = _mm512_sub_epi32(self.a, _mm512_rol_epi32(self.b, 27));
        self.a = _mm512_xor_si512(self.b, _mm512_rol_epi32(self.c, 17));
        self.b = _mm512_add_epi32(self.c, self.d);
        self.c = _mm512_add_epi32(self.d, e);
        self.d = _mm512_add_epi32(e, self.a);
        self.d
    }

    #[inline]
    #[target_feature(enable = "avx512f")]
    pub unsafe fn nextf_vec_scaled(&mut self, scale: __m512) -> __m512 {
        let v_u32 = unsafe { self.nextu_vec() };
        let v_f32 = _mm512_cvtepu32_ps(v_u32);
        _mm512_mul_ps(v_f32, scale)
    }

    #[inline]
    #[target_feature(enable = "avx512f")]
    pub unsafe fn randi_vec(&mut self, v_range: __m512i, v_min: __m512i) -> __m512i {
        const MERGE_MASK: u16 = 0xAAAA;
        let v_u32 = unsafe { self.nextu_vec() };
        let prod_even = _mm512_mul_epu32(v_u32, v_range);
        let res_even = _mm512_srli_epi64(prod_even, 32);
        let v_u32_shifted = _mm512_srli_epi64(v_u32, 32);
        let prod_odd = _mm512_mul_epu32(v_u32_shifted, v_range);
        let merged = _mm512_mask_blend_epi32(MERGE_MASK, res_even, prod_odd);
        _mm512_add_epi32(merged, v_min)
    }

    #[inline]
    #[target_feature(enable = "avx512f")]
    pub unsafe fn randf_vec(&mut self, v_mult: __m512, v_min: __m512) -> __m512 {
        let v_u32 = unsafe { self.nextu_vec() };
        let v_f32 = _mm512_cvtepu32_ps(v_u32);
        _mm512_add_ps(_mm512_mul_ps(v_f32, v_mult), v_min)
    }

    #[inline(always)]
    pub fn nextu(&mut self) -> [u32; JSF32X16] {
        unsafe {
            let mut result = [0u32; JSF32X16];
            _mm512_storeu_si512(result.as_mut_ptr() as *mut __m512i, self.nextu_vec());
            result
        }
    }

    pub fn nextf(&mut self) -> [f32; JSF32X16] {
        self.nextu().map(|x| x as f32 * FSCALE32)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::{safe_test, unsafe_test};

    safe_test!(Jsf32);
    unsafe_test!(Jsf32x16);
}