base64_simd/
encode.rs

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
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
use crate::{Config, Kind};
use crate::{STANDARD_CHARSET, URL_SAFE_CHARSET};

use vsimd::isa::{NEON, SSE2, WASM128};
use vsimd::tools::{read, write};
use vsimd::vector::{V128, V256};
use vsimd::{matches_isa, POD};
use vsimd::{Scalable, SIMD128, SIMD256};

#[inline(always)]
pub(crate) const fn encoded_length_unchecked(len: usize, config: Config) -> usize {
    let extra = len % 3;
    if extra == 0 {
        len / 3 * 4
    } else if config.extra.padding() {
        len / 3 * 4 + 4
    } else {
        len / 3 * 4 + extra + 1
    }
}

#[inline(always)]
unsafe fn encode_bits24(src: *const u8, dst: *mut u8, charset: *const u8) {
    let x = u32::from_be_bytes([0, read(src, 0), read(src, 1), read(src, 2)]);
    let mut i = 0;
    while i < 4 {
        let bits = (x >> (18 - i * 6)) & 0x3f;
        let y = read(charset, bits as usize);
        write(dst, i, y);
        i += 1;
    }
}

#[inline(always)]
unsafe fn encode_bits48(src: *const u8, dst: *mut u8, charset: *const u8) {
    let x = u64::from_be_bytes(src.cast::<[u8; 8]>().read());
    let mut i = 0;
    while i < 8 {
        let bits = (x >> (58 - i * 6)) & 0x3f;
        let y = read(charset, bits as usize);
        write(dst, i, y);
        i += 1;
    }
}

#[inline(always)]
unsafe fn encode_extra(extra: usize, src: *const u8, dst: *mut u8, charset: *const u8, padding: bool) {
    match extra {
        0 => {}
        1 => {
            let x = read(src, 0);
            let y1 = read(charset, (x >> 2) as usize);
            let y2 = read(charset, ((x << 6) >> 2) as usize);
            write(dst, 0, y1);
            write(dst, 1, y2);
            if padding {
                write(dst, 2, b'=');
                write(dst, 3, b'=');
            }
        }
        2 => {
            let x1 = read(src, 0);
            let x2 = read(src, 1);
            let y1 = read(charset, (x1 >> 2) as usize);
            let y2 = read(charset, (((x1 << 6) >> 2) | (x2 >> 4)) as usize);
            let y3 = read(charset, ((x2 << 4) >> 2) as usize);
            write(dst, 0, y1);
            write(dst, 1, y2);
            write(dst, 2, y3);
            if padding {
                write(dst, 3, b'=');
            }
        }
        _ => core::hint::unreachable_unchecked(),
    }
}

#[inline]
pub(crate) unsafe fn encode_fallback(mut src: *const u8, mut len: usize, mut dst: *mut u8, config: Config) {
    let kind = config.kind;
    let padding = config.extra.padding();

    let charset = match kind {
        Kind::Standard => STANDARD_CHARSET.as_ptr(),
        Kind::UrlSafe => URL_SAFE_CHARSET.as_ptr(),
    };

    const L: usize = 4;
    while len >= L * 6 + 2 {
        let mut i = 0;
        while i < L {
            encode_bits48(src, dst, charset);
            src = src.add(6);
            dst = dst.add(8);
            i += 1;
        }
        len -= L * 6;
    }

    while len >= 6 + 2 {
        encode_bits48(src, dst, charset);
        src = src.add(6);
        dst = dst.add(8);
        len -= 6;
    }

    let end = src.add(len / 3 * 3);
    while src < end {
        encode_bits24(src, dst, charset);
        src = src.add(3);
        dst = dst.add(4);
    }
    len %= 3;

    encode_extra(len, src, dst, charset, padding);
}

#[inline(always)]
pub(crate) unsafe fn encode_simd<S: SIMD256>(
    s: S,
    mut src: *const u8,
    mut len: usize,
    mut dst: *mut u8,
    config: Config,
) {
    let kind = config.kind;

    if len >= (6 + 24 + 4) {
        let (charset, shift_lut) = match kind {
            Kind::Standard => (STANDARD_CHARSET.as_ptr(), STANDARD_ENCODING_SHIFT_X2),
            Kind::UrlSafe => (URL_SAFE_CHARSET.as_ptr(), URL_SAFE_ENCODING_SHIFT_X2),
        };

        for _ in 0..2 {
            encode_bits24(src, dst, charset);
            src = src.add(3);
            dst = dst.add(4);
            len -= 3;
        }

        while len >= (24 + 4) {
            let x = s.v256_load_unaligned(src.sub(4));
            let y = encode_bytes24(s, x, shift_lut);
            s.v256_store_unaligned(dst, y);
            src = src.add(24);
            dst = dst.add(32);
            len -= 24;
        }
    }

    if len >= 12 + 4 {
        let shift_lut = match kind {
            Kind::Standard => STANDARD_ENCODING_SHIFT,
            Kind::UrlSafe => URL_SAFE_ENCODING_SHIFT,
        };

        let x = s.v128_load_unaligned(src);
        let y = encode_bytes12(s, x, shift_lut);
        s.v128_store_unaligned(dst, y);
        src = src.add(12);
        dst = dst.add(16);
        len -= 12;
    }

    encode_fallback(src, len, dst, config);
}

const SPLIT_SHUFFLE: V256 = V256::from_bytes([
    0x05, 0x04, 0x06, 0x05, 0x08, 0x07, 0x09, 0x08, //
    0x0b, 0x0a, 0x0c, 0x0b, 0x0e, 0x0d, 0x0f, 0x0e, //
    0x01, 0x00, 0x02, 0x01, 0x04, 0x03, 0x05, 0x04, //
    0x07, 0x06, 0x08, 0x07, 0x0a, 0x09, 0x0b, 0x0a, //
]);

#[inline(always)]
fn split_bits_x2<S: SIMD256>(s: S, x: V256) -> V256 {
    // x: {????|AAAB|BBCC|CDDD|EEEF|FFGG|GHHH|????}

    let x0 = s.u8x16x2_swizzle(x, SPLIT_SHUFFLE);
    // x0: {bbbbcccc|aaaaaabb|ccdddddd|bbbbcccc} x8 (1021)

    if matches_isa!(S, SSE2) {
        let m1 = s.u32x8_splat(u32::from_le_bytes([0x00, 0xfc, 0xc0, 0x0f]));
        let x1 = s.v256_and(x0, m1);
        // x1: {00000000|aaaaaa00|cc000000|0000cccc} x8

        let m2 = s.u32x8_splat(u32::from_le_bytes([0xf0, 0x03, 0x3f, 0x00]));
        let x2 = s.v256_and(x0, m2);
        // x2: {bbbb0000|000000bb|00dddddd|00000000} x8

        let m3 = s.u32x8_splat(u32::from_le_bytes([0x40, 0x00, 0x00, 0x04]));
        let x3 = s.u16x16_mul_hi(x1, m3);
        // x3: {00aaaaaa|00000000|00cccccc|00000000} x8

        let m4 = s.u32x8_splat(u32::from_le_bytes([0x10, 0x00, 0x00, 0x01]));
        let x4 = s.i16x16_mul_lo(x2, m4);
        // x4: {00000000|00bbbbbb|00000000|00dddddd} x8

        return s.v256_or(x3, x4);
        // {00aaaaaa|00bbbbbb|00cccccc|00dddddd} x8
    }

    if matches_isa!(S, NEON | WASM128) {
        let m1 = s.u32x8_splat(u32::from_le_bytes([0x00, 0xfc, 0x00, 0x00]));
        let x1 = s.u16x16_shr::<10>(s.v256_and(x0, m1));
        // x1: {00aaaaaa|000000000|00000000|00000000} x8

        let m2 = s.u32x8_splat(u32::from_le_bytes([0xf0, 0x03, 0x00, 0x00]));
        let x2 = s.u16x16_shl::<4>(s.v256_and(x0, m2));
        // x2: {00000000|00bbbbbb|00000000|00000000} x8

        let m3 = s.u32x8_splat(u32::from_le_bytes([0x00, 0x00, 0xc0, 0x0f]));
        let x3 = s.u16x16_shr::<6>(s.v256_and(x0, m3));
        // x3: {00000000|00000000|00cccccc|00000000} x8

        let m4 = s.u32x8_splat(u32::from_le_bytes([0x00, 0x00, 0x3f, 0x00]));
        let x4 = s.u16x16_shl::<8>(s.v256_and(x0, m4));
        // x4: {00000000|00000000|00000000|00dddddd} x8

        return s.v256_or(s.v256_or(x1, x2), s.v256_or(x3, x4));
        // {00aaaaaa|00bbbbbb|00cccccc|00dddddd} x8
    }

    unreachable!()
}

#[inline(always)]
fn split_bits_x1<S: SIMD128>(s: S, x: V128) -> V128 {
    // x: {AAAB|BBCC|CDDD|????}

    const SHUFFLE: V128 = SPLIT_SHUFFLE.to_v128x2().1;
    let x0 = s.u8x16_swizzle(x, SHUFFLE);
    // x0: {bbbbcccc|aaaaaabb|ccdddddd|bbbbcccc} x8 (1021)

    if matches_isa!(S, SSE2) {
        let m1 = s.u32x4_splat(u32::from_le_bytes([0x00, 0xfc, 0xc0, 0x0f]));
        let x1 = s.v128_and(x0, m1);

        let m2 = s.u32x4_splat(u32::from_le_bytes([0xf0, 0x03, 0x3f, 0x00]));
        let x2 = s.v128_and(x0, m2);

        let m3 = s.u32x4_splat(u32::from_le_bytes([0x40, 0x00, 0x00, 0x04]));
        let x3 = s.u16x8_mul_hi(x1, m3);

        let m4 = s.u32x4_splat(u32::from_le_bytes([0x10, 0x00, 0x00, 0x01]));
        let x4 = s.i16x8_mul_lo(x2, m4);

        return s.v128_or(x3, x4);
    }

    if matches_isa!(S, NEON | WASM128) {
        let m1 = s.u32x4_splat(u32::from_le_bytes([0x00, 0xfc, 0x00, 0x00]));
        let x1 = s.u16x8_shr::<10>(s.v128_and(x0, m1));

        let m2 = s.u32x4_splat(u32::from_le_bytes([0xf0, 0x03, 0x00, 0x00]));
        let x2 = s.u16x8_shl::<4>(s.v128_and(x0, m2));

        let m3 = s.u32x4_splat(u32::from_le_bytes([0x00, 0x00, 0xc0, 0x0f]));
        let x3 = s.u16x8_shr::<6>(s.v128_and(x0, m3));

        let m4 = s.u32x4_splat(u32::from_le_bytes([0x00, 0x00, 0x3f, 0x00]));
        let x4 = s.u16x8_shl::<8>(s.v128_and(x0, m4));

        return s.v128_or(s.v128_or(x1, x2), s.v128_or(x3, x4));
    }

    unreachable!()
}

#[inline]
const fn encoding_shift(charset: &'static [u8; 64]) -> V128 {
    // 0~25     'A'   [13]
    // 26~51    'a'   [0]
    // 52~61    '0'   [1~10]
    // 62       c62   [11]
    // 63       c63   [12]

    let mut lut = [0x80; 16];
    lut[13] = b'A';
    lut[0] = b'a' - 26;
    let mut i = 1;
    while i <= 10 {
        lut[i] = b'0'.wrapping_sub(52);
        i += 1;
    }
    lut[11] = charset[62].wrapping_sub(62);
    lut[12] = charset[63].wrapping_sub(63);
    V128::from_bytes(lut)
}

const STANDARD_ENCODING_SHIFT: V128 = encoding_shift(STANDARD_CHARSET);
const URL_SAFE_ENCODING_SHIFT: V128 = encoding_shift(URL_SAFE_CHARSET);

const STANDARD_ENCODING_SHIFT_X2: V256 = STANDARD_ENCODING_SHIFT.x2();
const URL_SAFE_ENCODING_SHIFT_X2: V256 = URL_SAFE_ENCODING_SHIFT.x2();

#[inline(always)]
fn encode_values<S: Scalable<V>, V: POD>(s: S, x: V, shift_lut: V) -> V {
    // x: {00aaaaaa|00bbbbbb|00cccccc|00dddddd} xn

    let x1 = s.u8xn_sub_sat(x, s.u8xn_splat(51));
    // 0~25    => 0
    // 26~51   => 0
    // 52~61   => 1~10
    // 62      => 11
    // 63      => 12

    let x2 = s.i8xn_lt(x, s.u8xn_splat(26));
    let x3 = s.and(x2, s.u8xn_splat(13));
    let x4 = s.or(x1, x3);
    // 0~25    => 0xff  => 13
    // 26~51   => 0     => 0
    // 52~61   => 0     => 1~10
    // 62      => 0     => 11
    // 63      => 0     => 12

    let shift = s.u8x16xn_swizzle(shift_lut, x4);
    s.u8xn_add(x, shift)
    // {{ascii}} xn
}

#[inline(always)]
fn encode_bytes24<S: SIMD256>(s: S, x: V256, shift_lut: V256) -> V256 {
    // x: {????|AAAB|BBCC|CDDD|EEEF|FFGG|GHHH|????}

    let values = split_bits_x2(s, x);
    // values: {00aaaaaa|00bbbbbb|00cccccc|00dddddd} x8

    encode_values(s, values, shift_lut)
    // {{ascii}} x32
}

#[inline(always)]
fn encode_bytes12<S: SIMD256>(s: S, x: V128, shift_lut: V128) -> V128 {
    // x: {AAAB|BBCC|CDDD|????}

    let values = split_bits_x1(s, x);
    // values: {00aaaaaa|00bbbbbb|00cccccc|00dddddd} x4

    encode_values(s, values, shift_lut)
    // {{ascii}} x16
}