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
|
#include "utf8.h"
#include <stddef.h>
#include <stdio.h>
#include <stdint.h>
#include <immintrin.h>
#define BIT(B,P) ((B >> (P-1)) & 0x1)
void _print_mmy(const char * msg, __m256i chunk)
{
printf("%s:", msg);
// unpack the first 8 bytes, padding with zeros
uint64_t a = _mm256_extract_epi64(chunk, 0);
uint64_t b = _mm256_extract_epi64(chunk, 1);
uint64_t c = _mm256_extract_epi64(chunk, 2);
uint64_t d = _mm256_extract_epi64(chunk, 3);
printf("%.2x%.2x%.2x%.2x %.2x%.2x%.2x%.2x %.2x%.2x%.2x%.2x %.2x%.2x%.2x%.2x "
"%.2x%.2x%.2x%.2x %.2x%.2x%.2x%.2x %.2x%.2x%.2x%.2x %.2x%.2x%.2x%.2x",
(unsigned char)((a >> 0) & 0xff),
(unsigned char)((a >> 8) & 0xff),
(unsigned char)((a >> 16) & 0xff),
(unsigned char)((a >> 24) & 0xff),
(unsigned char)((a >> 32) & 0xff),
(unsigned char)((a >> 40) & 0xff),
(unsigned char)((a >> 48) & 0xff),
(unsigned char)((a >> 56) & 0xff),
(unsigned char)((b >> 0) & 0xff),
(unsigned char)((b >> 8) & 0xff),
(unsigned char)((b >> 16) & 0xff),
(unsigned char)((b >> 24) & 0xff),
(unsigned char)((b >> 32) & 0xff),
(unsigned char)((b >> 40) & 0xff),
(unsigned char)((b >> 48) & 0xff),
(unsigned char)((b >> 56) & 0xff),
(unsigned char)((c >> 0) & 0xff),
(unsigned char)((c >> 8) & 0xff),
(unsigned char)((c >> 16) & 0xff),
(unsigned char)((c >> 24) & 0xff),
(unsigned char)((c >> 32) & 0xff),
(unsigned char)((c >> 40) & 0xff),
(unsigned char)((c >> 48) & 0xff),
(unsigned char)((c >> 56) & 0xff),
(unsigned char)((d >> 0) & 0xff),
(unsigned char)((d >> 8) & 0xff),
(unsigned char)((d >> 16) & 0xff),
(unsigned char)((d >> 24) & 0xff),
(unsigned char)((d >> 32) & 0xff),
(unsigned char)((d >> 40) & 0xff),
(unsigned char)((d >> 48) & 0xff),
(unsigned char)((d >> 56) & 0xff)
);
printf("\n");
}
ssize_t fu8_count_utf8_codepoints_avx(const char * utf8, size_t len)
{
const uint8_t * encoded = (const uint8_t*)utf8;
__builtin_prefetch(encoded, 0, 0);
size_t num_codepoints = 0;
__m256i chunk;
if (len == 0) {
return 0;
}
__m256i zero = _mm256_set1_epi8(0x00);
while (len >= 32) {
chunk = _mm256_loadu_si256((__m256i*)encoded);
if (_mm256_movemask_epi8(chunk) == 0) {
// valid ascii chars!
len -= 32;
encoded += 32;
num_codepoints += 32;
continue;
}
__builtin_prefetch(encoded+32, 0, 0);
__m256i count = _mm256_set1_epi8(0x1);
//_print_mm256x("chunk", chunk);
// fight against the fact that there is no comparison on unsigned values
__m256i chunk_signed = _mm256_add_epi8(chunk, _mm256_set1_epi8(0x80));
//_print_mm256x("shunk", chunk_signed);
// ERROR checking
// checking procedure works the following way:
//
// 1) mark all continuation bytes with either 0x1, 0x3, 0x7 (one, two or three bytes continuation)
// 2) then check that there is no byte that has an invalid continuation
__m256i twobytemarker = _mm256_cmpgt_epi8( chunk_signed, _mm256_set1_epi8(0xc0-1-0x80));
__m256i threebytemarker = _mm256_cmpgt_epi8(chunk_signed, _mm256_set1_epi8(0xe0-1-0x80));
__m256i fourbytemarker = _mm256_cmpgt_epi8( chunk_signed, _mm256_set1_epi8(0xf0-1-0x80));
// the general idea of the following code collects 0xff for each byte position
// in the variable contbytes.
// at the end check if each position in contbytes set to 0xff is a valid continuation byte
// check that 0xc0 > 0xc2
__m256i validtwobm = _mm256_cmpgt_epi8(chunk_signed, _mm256_set1_epi8(0xc2-1-0x80));
if (_mm256_movemask_epi8(_mm256_xor_si256(validtwobm, twobytemarker)) != 0) {
// two byte marker should not be in range [0xc0-0xc2)
return -1;
}
__m256i state2 = _mm256_andnot_si256(threebytemarker, twobytemarker);
__m256i contbytes = _mm256_slli_si256(_mm256_blendv_epi8(state2, _mm256_set1_epi8(0x1), twobytemarker), 1);
if (_mm256_movemask_epi8(threebytemarker) != 0) {
// contains at least one 3 byte marker
__m256i istate3 = _mm256_andnot_si256(fourbytemarker, threebytemarker);
__m256i state3 = _mm256_slli_si256(_mm256_blendv_epi8(zero, _mm256_set1_epi8(0x3), istate3), 1);
state3 = _mm256_or_si256(state3, _mm256_slli_si256(state3, 1));
contbytes = _mm256_or_si256(contbytes, state3);
// range check
__m256i equal_e0 = _mm256_cmpeq_epi8(_mm256_blendv_epi8(zero, chunk_signed, istate3),
_mm256_set1_epi8(0xe0-0x80));
if (_mm256_movemask_epi8(equal_e0) != 0) {
__m256i mask = _mm256_blendv_epi8(_mm256_set1_epi8(0x7f), chunk_signed, _mm256_slli_si256(equal_e0, 1));
__m256i check_surrogate = _mm256_cmpgt_epi8(_mm256_set1_epi8(0xa0-0x80), mask); // lt
if (_mm256_movemask_epi8(check_surrogate) != 0) {
// invalid surrograte character!!!
return -1;
}
}
// verify that there are now surrogates
if (!ALLOW_SURROGATES) {
__m256i equal_ed = _mm256_cmpeq_epi8(_mm256_blendv_epi8(zero, chunk_signed, istate3),
_mm256_set1_epi8(0xed-0x80));
if (_mm256_movemask_epi8(equal_ed) != 0) {
__m256i mask = _mm256_blendv_epi8(_mm256_set1_epi8(0x80), chunk_signed, _mm256_slli_si256(equal_ed, 1));
__m256i check_surrogate = _mm256_cmpgt_epi8(mask, _mm256_set1_epi8(0xa0-1-0x80));
if (_mm256_movemask_epi8(check_surrogate) != 0) {
// invalid surrograte character!!!
return -1;
}
}
}
}
if (_mm256_movemask_epi8(fourbytemarker) != 0) {
// contain a 4 byte marker
__m256i istate4 = _mm256_slli_si256(_mm256_blendv_epi8(zero, _mm256_set1_epi8(0x7), fourbytemarker), 1);
__m256i state4 =_mm256_or_si256(istate4, _mm256_slli_si256(istate4, 1));
state4 =_mm256_or_si256(state4, _mm256_slli_si256(istate4, 2));
contbytes = _mm256_or_si256(contbytes, state4);
// range check, filter out f0 and
__m256i equal_f0 = _mm256_cmpeq_epi8(_mm256_blendv_epi8(zero, chunk_signed, fourbytemarker),
_mm256_set1_epi8(0xf0-0x80));
if (_mm256_movemask_epi8(equal_f0) != 0) {
__m256i mask = _mm256_blendv_epi8(_mm256_set1_epi8(0x7f), chunk_signed, _mm256_slli_si256(equal_f0, 1));
__m256i check_surrogate = _mm256_cmpgt_epi8(_mm256_set1_epi8(0x90-0x80), mask);
if (_mm256_movemask_epi8(check_surrogate) != 0) {
return -1;
}
}
__m256i equal_f4 = _mm256_cmpeq_epi8(_mm256_blendv_epi8(zero, chunk_signed, fourbytemarker),
_mm256_set1_epi8(0xf4-0x80));
if (_mm256_movemask_epi8(equal_f4) != 0) {
__m256i mask = _mm256_blendv_epi8(_mm256_set1_epi8(0x80), chunk_signed, _mm256_slli_si256(equal_f4, 1));
__m256i check_surrogate = _mm256_cmpgt_epi8(mask, _mm256_set1_epi8(0x90-1-0x80));
if (_mm256_movemask_epi8(check_surrogate) != 0) {
return -1;
}
}
__m256i equal_f5_gt = _mm256_cmpgt_epi8(_mm256_blendv_epi8(zero, chunk_signed, fourbytemarker),
_mm256_set1_epi8(0xf4-0x80));
if (_mm256_movemask_epi8(equal_f5_gt) != 0) {
return -1;
}
}
// now check that contbytes and the actual byte values have a valid
// continuation at each position the marker indicates to have one
__m256i check_cont = _mm256_cmpgt_epi8(contbytes, zero);
__m256i contpos = _mm256_and_si256(_mm256_set1_epi8(0xc0), chunk);
contpos = _mm256_cmpeq_epi8(_mm256_set1_epi8(0x80), contpos);
__m256i validcont = _mm256_xor_si256(check_cont, contpos);
if (_mm256_movemask_epi8(validcont) != 0) {
// uff, nope, that is really not utf8
return -1;
}
// CORRECT, calculate the length
// copy 0x00 over to each place which is a continuation byte
count = _mm256_blendv_epi8(count, zero, contpos);
// count the code points using 2x 32 bit hadd and one last 16 hadd
// the result will end up at the lowest position
count = _mm256_hadd_epi32(count, zero);
count = _mm256_hadd_epi32(count, zero);
count = _mm256_hadd_epi16(count, zero);
uint16_t c = _mm256_extract_epi16(count, 0);
uint16_t c2 = _mm256_extract_epi16(count, 8);
uint16_t points = (c & 0xff) + ((c >> 8) & 0xff) + (c2 & 0xff) + ((c2 >> 8) & 0xff);
// these cases need to be handled:
// 16 byte boundary -> | <- 16 byte boundary
// -----------------------------------------+--------------------
// 1) 2 byte code point. e.g. ... c2 | 80 ...
// 2) 3 byte code point. e.g. ... e6 | 80 80 ...
// 3) 3 byte code point. e.g. ... e6 80 | 80 ...
// 4) 4 byte code point. e.g. ... f2 | 80 80 80 ...
// 5) 4 byte code point. e.g. ... f2 80 | 80 80 ...
// 6) 4 byte code point. e.g. ... f2 80 80 | 80 ...
//
int mask_chunk = _mm256_movemask_epi8(chunk);
int mask_conti = _mm256_movemask_epi8(contpos);
// little endian
int lenoff = 32;
int minus_codepoints = 0;
if (BIT(mask_chunk, 32) != 0 && BIT(mask_conti, 32) == 0) { // 1), 2), 4)
minus_codepoints = 1;
lenoff -= 1;
} else if (BIT(mask_chunk, 31) != 0 && BIT(mask_conti, 31) == 0 &&
BIT(mask_conti, 32) == 1) { // 3), 5)
minus_codepoints = 1;
lenoff -= 2;
} else if (BIT(mask_chunk, 30) != 0 && BIT(mask_conti, 30) == 0 &&
BIT(mask_conti, 31) == 1 && BIT(mask_conti, 32) == 1) { // 6)
minus_codepoints = 1;
lenoff -= 3;
}
num_codepoints += points - minus_codepoints;
len -= lenoff;
encoded += lenoff;
}
if (len == 0) {
return num_codepoints;
}
ssize_t result = fu8_count_utf8_codepoints_seq(encoded, len);
if (result == -1) {
return -1;
}
return num_codepoints + result;
return -1;
}
|
