urng 0.4.5

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
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

use crate::rng64::Philox64;
use rayon::prelude::*;
use std::slice::from_raw_parts_mut;

/// Creates a new heap-allocated `Philox64` and returns a raw pointer to it.
/// The caller is responsible for freeing it with [`philox64_free`].
#[unsafe(no_mangle)]
pub extern "C" fn philox64_new(seed: u64) -> *mut Philox64 {
    Box::into_raw(Box::new(Philox64::new(seed)))
}
/// Frees a `Philox64` instance previously created by [`philox64_new`].
/// Does nothing if `ptr` is null.
#[unsafe(no_mangle)]
pub extern "C" fn philox64_free(ptr: *mut Philox64) {
    if !ptr.is_null() {
        unsafe { drop(Box::from_raw(ptr)) }
    }
}

/// Parallel batch size for Philox64 (elements per thread task).
const PHILOX64_PAR_CHUNK: usize = 4096;

/// Fills `out[0..count]` with raw `u64` random values using parallel counter-based generation.
#[unsafe(no_mangle)]
pub extern "C" fn philox64_next_u64s(ptr: *mut Philox64, out: *mut u64, count: usize) {
    unsafe {
        let rng = &mut *ptr;
        let buffer = from_raw_parts_mut(out, count);
        let c0 = rng.c;
        let k = rng.k;

        buffer
            .par_chunks_mut(PHILOX64_PAR_CHUNK)
            .enumerate()
            .for_each(|(chunk_idx, chunk)| {
                let chunk_base_block = (chunk_idx * PHILOX64_PAR_CHUNK) / 2;
                let mut chunks_exact = chunk.chunks_exact_mut(2);
                let mut b_offset = 0u64;

                for dst in chunks_exact.by_ref() {
                    let mut c = c0;
                    let (new_c0, overflow) =
                        c[0].overflowing_add((chunk_base_block as u64) + b_offset);
                    c[0] = new_c0;
                    if overflow {
                        c[1] = c[1].wrapping_add(1);
                    }

                    let result = Philox64::compute(c, k);
                    dst[0] = result[0];
                    dst[1] = result[1];
                    b_offset += 1;
                }

                let rem = chunks_exact.into_remainder();
                if !rem.is_empty() {
                    let mut c = c0;
                    let (new_c0, overflow) =
                        c[0].overflowing_add((chunk_base_block as u64) + b_offset);
                    c[0] = new_c0;
                    if overflow {
                        c[1] = c[1].wrapping_add(1);
                    }

                    let result = Philox64::compute(c, k);
                    for j in 0..rem.len() {
                        rem[j] = result[j];
                    }
                }
            });

        let num_blocks = ((count + 1) / 2) as u64;
        let (new_c0, carry) = rng.c[0].overflowing_add(num_blocks);
        rng.c[0] = new_c0;
        if carry {
            rng.c[1] = rng.c[1].wrapping_add(1);
        }
    }
}

/// Fills `out[0..count]` with `f64` values in `[0, 1)` using parallel counter-based generation.
#[unsafe(no_mangle)]
pub extern "C" fn philox64_next_f64s(ptr: *mut Philox64, out: *mut f64, count: usize) {
    unsafe {
        let rng = &mut *ptr;
        let buffer = from_raw_parts_mut(out, count);
        let c0 = rng.c;
        let k = rng.k;
        let scale = 1.0f64 / (u64::MAX as f64 + 1.0);

        buffer
            .par_chunks_mut(PHILOX64_PAR_CHUNK)
            .enumerate()
            .for_each(|(chunk_idx, chunk)| {
                let chunk_base_block = (chunk_idx * PHILOX64_PAR_CHUNK) / 2;
                let mut chunks_exact = chunk.chunks_exact_mut(2);
                let mut b_offset = 0u64;

                for dst in chunks_exact.by_ref() {
                    let mut c = c0;
                    let (new_c0, overflow) =
                        c[0].overflowing_add((chunk_base_block as u64) + b_offset);
                    c[0] = new_c0;
                    if overflow {
                        c[1] = c[1].wrapping_add(1);
                    }

                    let result = Philox64::compute(c, k);
                    dst[0] = result[0] as f64 * scale;
                    dst[1] = result[1] as f64 * scale;
                    b_offset += 1;
                }

                let rem = chunks_exact.into_remainder();
                if !rem.is_empty() {
                    let mut c = c0;
                    let (new_c0, overflow) =
                        c[0].overflowing_add((chunk_base_block as u64) + b_offset);
                    c[0] = new_c0;
                    if overflow {
                        c[1] = c[1].wrapping_add(1);
                    }

                    let result = Philox64::compute(c, k);
                    for j in 0..rem.len() {
                        rem[j] = result[j] as f64 * scale;
                    }
                }
            });

        let num_blocks = ((count + 1) / 2) as u64;
        let (new_c0, carry) = rng.c[0].overflowing_add(num_blocks);
        rng.c[0] = new_c0;
        if carry {
            rng.c[1] = rng.c[1].wrapping_add(1);
        }
    }
}
/// Fills `out[0..count]` with `i64` values in `[min, max]` using parallel counter-based generation.
#[unsafe(no_mangle)]
pub extern "C" fn philox64_rand_i64s(
    ptr: *mut Philox64,
    out: *mut i64,
    count: usize,
    min: i64,
    max: i64,
) {
    unsafe {
        let rng = &mut *ptr;
        let buffer = from_raw_parts_mut(out, count);
        let c0 = rng.c;
        let k = rng.k;
        let range = (max as i128 - min as i128 + 1) as u128;

        buffer
            .par_chunks_mut(PHILOX64_PAR_CHUNK)
            .enumerate()
            .for_each(|(chunk_idx, chunk)| {
                let chunk_base_block = (chunk_idx * PHILOX64_PAR_CHUNK) / 2;
                let mut chunks_exact = chunk.chunks_exact_mut(2);
                let mut b_offset = 0u64;

                for dst in chunks_exact.by_ref() {
                    let mut c = c0;
                    let (new_c0, overflow) =
                        c[0].overflowing_add((chunk_base_block as u64) + b_offset);
                    c[0] = new_c0;
                    if overflow {
                        c[1] = c[1].wrapping_add(1);
                    }

                    let result = Philox64::compute(c, k);
                    dst[0] = ((result[0] as u128 * range) >> 64) as i64 + min;
                    dst[1] = ((result[1] as u128 * range) >> 64) as i64 + min;
                    b_offset += 1;
                }

                let rem = chunks_exact.into_remainder();
                if !rem.is_empty() {
                    let mut c = c0;
                    let (new_c0, overflow) =
                        c[0].overflowing_add((chunk_base_block as u64) + b_offset);
                    c[0] = new_c0;
                    if overflow {
                        c[1] = c[1].wrapping_add(1);
                    }

                    let result = Philox64::compute(c, k);
                    for j in 0..rem.len() {
                        rem[j] = ((result[j] as u128 * range) >> 64) as i64 + min;
                    }
                }
            });

        let num_blocks = ((count + 1) / 2) as u64;
        let (new_c0, carry) = rng.c[0].overflowing_add(num_blocks);
        rng.c[0] = new_c0;
        if carry {
            rng.c[1] = rng.c[1].wrapping_add(1);
        }
    }
}
/// Fills `out[0..count]` with `f64` values in `[min, max)` using parallel counter-based generation.
#[unsafe(no_mangle)]
pub extern "C" fn philox64_rand_f64s(
    ptr: *mut Philox64,
    out: *mut f64,
    count: usize,
    min: f64,
    max: f64,
) {
    unsafe {
        let rng = &mut *ptr;
        let buffer = from_raw_parts_mut(out, count);
        let c0 = rng.c;
        let k = rng.k;
        let scale_val = 1.0f64 / (u64::MAX as f64 + 1.0);
        let range_val = max - min;

        buffer
            .par_chunks_mut(PHILOX64_PAR_CHUNK)
            .enumerate()
            .for_each(|(chunk_idx, chunk)| {
                let chunk_base_block = (chunk_idx * PHILOX64_PAR_CHUNK) / 2;
                let mut chunks_exact = chunk.chunks_exact_mut(2);
                let mut b_offset = 0u64;

                for dst in chunks_exact.by_ref() {
                    let mut c = c0;
                    let (new_c0, overflow) =
                        c[0].overflowing_add((chunk_base_block as u64) + b_offset);
                    c[0] = new_c0;
                    if overflow {
                        c[1] = c[1].wrapping_add(1);
                    }

                    let result = Philox64::compute(c, k);
                    dst[0] = (result[0] as f64 * scale_val) * range_val + min;
                    dst[1] = (result[1] as f64 * scale_val) * range_val + min;
                    b_offset += 1;
                }

                let rem = chunks_exact.into_remainder();
                if !rem.is_empty() {
                    let mut c = c0;
                    let (new_c0, overflow) =
                        c[0].overflowing_add((chunk_base_block as u64) + b_offset);
                    c[0] = new_c0;
                    if overflow {
                        c[1] = c[1].wrapping_add(1);
                    }

                    let result = Philox64::compute(c, k);
                    for j in 0..rem.len() {
                        rem[j] = (result[j] as f64 * scale_val) * range_val + min;
                    }
                }
            });

        let num_blocks = ((count + 1) / 2) as u64;
        let (new_c0, carry) = rng.c[0].overflowing_add(num_blocks);
        rng.c[0] = new_c0;
        if carry {
            rng.c[1] = rng.c[1].wrapping_add(1);
        }
    }
}