/src/cpython3/Objects/stringlib/find_max_char.h
Line | Count | Source |
1 | | /* Finding the optimal width of unicode characters in a buffer */ |
2 | | |
3 | | /* find_max_char for one-byte will work for bytes objects as well. */ |
4 | | #if !STRINGLIB_IS_UNICODE && STRINGLIB_SIZEOF_CHAR > 1 |
5 | | # error "find_max_char.h is specific to Unicode" |
6 | | #endif |
7 | | |
8 | | /* Mask to quickly check whether a C 'size_t' contains a |
9 | | non-ASCII, UTF8-encoded char. */ |
10 | | #if (SIZEOF_SIZE_T == 8) |
11 | 2.45M | # define UCS1_ASCII_CHAR_MASK 0x8080808080808080ULL |
12 | | #elif (SIZEOF_SIZE_T == 4) |
13 | | # define UCS1_ASCII_CHAR_MASK 0x80808080U |
14 | | #else |
15 | | # error C 'size_t' size should be either 4 or 8! |
16 | | #endif |
17 | | |
18 | | #if STRINGLIB_SIZEOF_CHAR == 1 |
19 | | |
20 | | Py_LOCAL_INLINE(Py_UCS4) |
21 | | STRINGLIB(find_max_char)(const STRINGLIB_CHAR *begin, const STRINGLIB_CHAR *end) |
22 | 721k | { |
23 | 721k | const unsigned char *p = (const unsigned char *) begin; |
24 | 721k | const unsigned char *_end = (const unsigned char *)end; |
25 | | |
26 | 4.79M | while (p < _end) { |
27 | 4.11M | if (_Py_IS_ALIGNED(p, ALIGNOF_SIZE_T)) { |
28 | | /* Help register allocation */ |
29 | 530k | const unsigned char *_p = p; |
30 | 2.97M | while (_p + SIZEOF_SIZE_T <= _end) { |
31 | 2.45M | size_t value = *(const size_t *) _p; |
32 | 2.45M | if (value & UCS1_ASCII_CHAR_MASK) |
33 | 10.4k | return 255; |
34 | 2.44M | _p += SIZEOF_SIZE_T; |
35 | 2.44M | } |
36 | 520k | p = _p; |
37 | 520k | if (p == _end) |
38 | 23.6k | break; |
39 | 520k | } |
40 | 4.07M | if (*p++ & 0x80) |
41 | 10.4k | return 255; |
42 | 4.07M | } |
43 | 700k | return 127; |
44 | 721k | } unicode_writer.c:ucs1lib_find_max_char Line | Count | Source | 22 | 318k | { | 23 | 318k | const unsigned char *p = (const unsigned char *) begin; | 24 | 318k | const unsigned char *_end = (const unsigned char *)end; | 25 | | | 26 | 1.98M | while (p < _end) { | 27 | 1.67M | if (_Py_IS_ALIGNED(p, ALIGNOF_SIZE_T)) { | 28 | | /* Help register allocation */ | 29 | 215k | const unsigned char *_p = p; | 30 | 637k | while (_p + SIZEOF_SIZE_T <= _end) { | 31 | 421k | size_t value = *(const size_t *) _p; | 32 | 421k | if (value & UCS1_ASCII_CHAR_MASK) | 33 | 0 | return 255; | 34 | 421k | _p += SIZEOF_SIZE_T; | 35 | 421k | } | 36 | 215k | p = _p; | 37 | 215k | if (p == _end) | 38 | 7.31k | break; | 39 | 215k | } | 40 | 1.66M | if (*p++ & 0x80) | 41 | 0 | return 255; | 42 | 1.66M | } | 43 | 318k | return 127; | 44 | 318k | } |
unicodeobject.c:ucs1lib_find_max_char Line | Count | Source | 22 | 402k | { | 23 | 402k | const unsigned char *p = (const unsigned char *) begin; | 24 | 402k | const unsigned char *_end = (const unsigned char *)end; | 25 | | | 26 | 2.80M | while (p < _end) { | 27 | 2.44M | if (_Py_IS_ALIGNED(p, ALIGNOF_SIZE_T)) { | 28 | | /* Help register allocation */ | 29 | 315k | const unsigned char *_p = p; | 30 | 2.33M | while (_p + SIZEOF_SIZE_T <= _end) { | 31 | 2.03M | size_t value = *(const size_t *) _p; | 32 | 2.03M | if (value & UCS1_ASCII_CHAR_MASK) | 33 | 10.4k | return 255; | 34 | 2.02M | _p += SIZEOF_SIZE_T; | 35 | 2.02M | } | 36 | 305k | p = _p; | 37 | 305k | if (p == _end) | 38 | 16.3k | break; | 39 | 305k | } | 40 | 2.41M | if (*p++ & 0x80) | 41 | 10.4k | return 255; | 42 | 2.41M | } | 43 | 381k | return 127; | 44 | 402k | } |
Unexecuted instantiation: unicodeobject.c:asciilib_find_max_char Unexecuted instantiation: bytes_methods.c:stringlib_find_max_char |
45 | | |
46 | | #undef ASCII_CHAR_MASK |
47 | | |
48 | | #else /* STRINGLIB_SIZEOF_CHAR == 1 */ |
49 | | |
50 | 1.44M | #define MASK_ASCII 0xFFFFFF80 |
51 | 738k | #define MASK_UCS1 0xFFFFFF00 |
52 | 706k | #define MASK_UCS2 0xFFFF0000 |
53 | | |
54 | 912k | #define MAX_CHAR_ASCII 0x7f |
55 | 402k | #define MAX_CHAR_UCS1 0xff |
56 | 466k | #define MAX_CHAR_UCS2 0xffff |
57 | 576k | #define MAX_CHAR_UCS4 0x10ffff |
58 | | |
59 | | Py_LOCAL_INLINE(Py_UCS4) |
60 | | STRINGLIB(find_max_char)(const STRINGLIB_CHAR *begin, const STRINGLIB_CHAR *end) |
61 | 912k | { |
62 | | #if STRINGLIB_SIZEOF_CHAR == 2 |
63 | 336k | const Py_UCS4 mask_limit = MASK_UCS1; |
64 | 336k | const Py_UCS4 max_char_limit = MAX_CHAR_UCS2; |
65 | | #elif STRINGLIB_SIZEOF_CHAR == 4 |
66 | 576k | const Py_UCS4 mask_limit = MASK_UCS2; |
67 | 576k | const Py_UCS4 max_char_limit = MAX_CHAR_UCS4; |
68 | | #else |
69 | | #error Invalid STRINGLIB_SIZEOF_CHAR (must be 1, 2 or 4) |
70 | | #endif |
71 | 912k | Py_UCS4 mask; |
72 | 912k | Py_ssize_t n = end - begin; |
73 | 912k | const STRINGLIB_CHAR *p = begin; |
74 | 912k | const STRINGLIB_CHAR *unrolled_end = begin + _Py_SIZE_ROUND_DOWN(n, 4); |
75 | 912k | Py_UCS4 max_char; |
76 | | |
77 | 912k | max_char = MAX_CHAR_ASCII; |
78 | 912k | mask = MASK_ASCII; |
79 | 13.5M | while (p < unrolled_end) { |
80 | 12.8M | STRINGLIB_CHAR bits = p[0] | p[1] | p[2] | p[3]; |
81 | 12.8M | if (bits & mask) { |
82 | 554k | if (mask == mask_limit) { |
83 | | /* Limit reached */ |
84 | 232k | return max_char_limit; |
85 | 232k | } |
86 | 322k | if (mask == MASK_ASCII) { |
87 | 268k | max_char = MAX_CHAR_UCS1; |
88 | 268k | mask = MASK_UCS1; |
89 | 268k | } |
90 | 54.0k | else { |
91 | | /* mask can't be MASK_UCS2 because of mask_limit above */ |
92 | 54.0k | assert(mask == MASK_UCS1); |
93 | 54.0k | max_char = MAX_CHAR_UCS2; |
94 | 54.0k | mask = MASK_UCS2; |
95 | 54.0k | } |
96 | | /* We check the new mask on the same chars in the next iteration */ |
97 | 322k | continue; |
98 | 322k | } |
99 | 12.3M | p += 4; |
100 | 12.3M | } |
101 | 2.04M | while (p < end) { |
102 | 1.41M | if (p[0] & mask) { |
103 | 267k | if (mask == mask_limit) { |
104 | | /* Limit reached */ |
105 | 56.8k | return max_char_limit; |
106 | 56.8k | } |
107 | 210k | if (mask == MASK_ASCII) { |
108 | 134k | max_char = MAX_CHAR_UCS1; |
109 | 134k | mask = MASK_UCS1; |
110 | 134k | } |
111 | 76.1k | else { |
112 | | /* mask can't be MASK_UCS2 because of mask_limit above */ |
113 | 76.1k | assert(mask == MASK_UCS1); |
114 | 76.1k | max_char = MAX_CHAR_UCS2; |
115 | 76.1k | mask = MASK_UCS2; |
116 | 76.1k | } |
117 | | /* We check the new mask on the same chars in the next iteration */ |
118 | 210k | continue; |
119 | 210k | } |
120 | 1.15M | p++; |
121 | 1.15M | } |
122 | 623k | return max_char; |
123 | 680k | } unicodeobject.c:ucs4lib_find_max_char Line | Count | Source | 61 | 576k | { | 62 | | #if STRINGLIB_SIZEOF_CHAR == 2 | 63 | | const Py_UCS4 mask_limit = MASK_UCS1; | 64 | | const Py_UCS4 max_char_limit = MAX_CHAR_UCS2; | 65 | | #elif STRINGLIB_SIZEOF_CHAR == 4 | 66 | 576k | const Py_UCS4 mask_limit = MASK_UCS2; | 67 | 576k | const Py_UCS4 max_char_limit = MAX_CHAR_UCS4; | 68 | | #else | 69 | | #error Invalid STRINGLIB_SIZEOF_CHAR (must be 1, 2 or 4) | 70 | | #endif | 71 | 576k | Py_UCS4 mask; | 72 | 576k | Py_ssize_t n = end - begin; | 73 | 576k | const STRINGLIB_CHAR *p = begin; | 74 | 576k | const STRINGLIB_CHAR *unrolled_end = begin + _Py_SIZE_ROUND_DOWN(n, 4); | 75 | 576k | Py_UCS4 max_char; | 76 | | | 77 | 576k | max_char = MAX_CHAR_ASCII; | 78 | 576k | mask = MASK_ASCII; | 79 | 10.9M | while (p < unrolled_end) { | 80 | 10.4M | STRINGLIB_CHAR bits = p[0] | p[1] | p[2] | p[3]; | 81 | 10.4M | if (bits & mask) { | 82 | 135k | if (mask == mask_limit) { | 83 | | /* Limit reached */ | 84 | 23.2k | return max_char_limit; | 85 | 23.2k | } | 86 | 111k | if (mask == MASK_ASCII) { | 87 | 57.9k | max_char = MAX_CHAR_UCS1; | 88 | 57.9k | mask = MASK_UCS1; | 89 | 57.9k | } | 90 | 54.0k | else { | 91 | | /* mask can't be MASK_UCS2 because of mask_limit above */ | 92 | 54.0k | assert(mask == MASK_UCS1); | 93 | 54.0k | max_char = MAX_CHAR_UCS2; | 94 | 54.0k | mask = MASK_UCS2; | 95 | 54.0k | } | 96 | | /* We check the new mask on the same chars in the next iteration */ | 97 | 111k | continue; | 98 | 111k | } | 99 | 10.2M | p += 4; | 100 | 10.2M | } | 101 | 1.72M | while (p < end) { | 102 | 1.17M | if (p[0] & mask) { | 103 | 162k | if (mask == mask_limit) { | 104 | | /* Limit reached */ | 105 | 5.29k | return max_char_limit; | 106 | 5.29k | } | 107 | 157k | if (mask == MASK_ASCII) { | 108 | 81.0k | max_char = MAX_CHAR_UCS1; | 109 | 81.0k | mask = MASK_UCS1; | 110 | 81.0k | } | 111 | 76.1k | else { | 112 | | /* mask can't be MASK_UCS2 because of mask_limit above */ | 113 | 76.1k | assert(mask == MASK_UCS1); | 114 | 76.1k | max_char = MAX_CHAR_UCS2; | 115 | 76.1k | mask = MASK_UCS2; | 116 | 76.1k | } | 117 | | /* We check the new mask on the same chars in the next iteration */ | 118 | 157k | continue; | 119 | 157k | } | 120 | 1.01M | p++; | 121 | 1.01M | } | 122 | 548k | return max_char; | 123 | 553k | } |
unicodeobject.c:ucs2lib_find_max_char Line | Count | Source | 61 | 336k | { | 62 | 336k | #if STRINGLIB_SIZEOF_CHAR == 2 | 63 | 336k | const Py_UCS4 mask_limit = MASK_UCS1; | 64 | 336k | const Py_UCS4 max_char_limit = MAX_CHAR_UCS2; | 65 | | #elif STRINGLIB_SIZEOF_CHAR == 4 | 66 | | const Py_UCS4 mask_limit = MASK_UCS2; | 67 | | const Py_UCS4 max_char_limit = MAX_CHAR_UCS4; | 68 | | #else | 69 | | #error Invalid STRINGLIB_SIZEOF_CHAR (must be 1, 2 or 4) | 70 | | #endif | 71 | 336k | Py_UCS4 mask; | 72 | 336k | Py_ssize_t n = end - begin; | 73 | 336k | const STRINGLIB_CHAR *p = begin; | 74 | 336k | const STRINGLIB_CHAR *unrolled_end = begin + _Py_SIZE_ROUND_DOWN(n, 4); | 75 | 336k | Py_UCS4 max_char; | 76 | | | 77 | 336k | max_char = MAX_CHAR_ASCII; | 78 | 336k | mask = MASK_ASCII; | 79 | 2.56M | while (p < unrolled_end) { | 80 | 2.43M | STRINGLIB_CHAR bits = p[0] | p[1] | p[2] | p[3]; | 81 | 2.43M | if (bits & mask) { | 82 | 419k | if (mask == mask_limit) { | 83 | | /* Limit reached */ | 84 | 209k | return max_char_limit; | 85 | 209k | } | 86 | 210k | if (mask == MASK_ASCII) { | 87 | 210k | max_char = MAX_CHAR_UCS1; | 88 | 210k | mask = MASK_UCS1; | 89 | 210k | } | 90 | 0 | else { | 91 | | /* mask can't be MASK_UCS2 because of mask_limit above */ | 92 | 0 | assert(mask == MASK_UCS1); | 93 | 0 | max_char = MAX_CHAR_UCS2; | 94 | 0 | mask = MASK_UCS2; | 95 | 0 | } | 96 | | /* We check the new mask on the same chars in the next iteration */ | 97 | 210k | continue; | 98 | 210k | } | 99 | 2.01M | p += 4; | 100 | 2.01M | } | 101 | 320k | while (p < end) { | 102 | 244k | if (p[0] & mask) { | 103 | 104k | if (mask == mask_limit) { | 104 | | /* Limit reached */ | 105 | 51.5k | return max_char_limit; | 106 | 51.5k | } | 107 | 53.3k | if (mask == MASK_ASCII) { | 108 | 53.3k | max_char = MAX_CHAR_UCS1; | 109 | 53.3k | mask = MASK_UCS1; | 110 | 53.3k | } | 111 | 0 | else { | 112 | | /* mask can't be MASK_UCS2 because of mask_limit above */ | 113 | 0 | assert(mask == MASK_UCS1); | 114 | 0 | max_char = MAX_CHAR_UCS2; | 115 | 0 | mask = MASK_UCS2; | 116 | 0 | } | 117 | | /* We check the new mask on the same chars in the next iteration */ | 118 | 53.3k | continue; | 119 | 53.3k | } | 120 | 139k | p++; | 121 | 139k | } | 122 | 75.2k | return max_char; | 123 | 126k | } |
|
124 | | |
125 | | #undef MASK_ASCII |
126 | | #undef MASK_UCS1 |
127 | | #undef MASK_UCS2 |
128 | | #undef MAX_CHAR_ASCII |
129 | | #undef MAX_CHAR_UCS1 |
130 | | #undef MAX_CHAR_UCS2 |
131 | | #undef MAX_CHAR_UCS4 |
132 | | |
133 | | #endif /* STRINGLIB_SIZEOF_CHAR == 1 */ |
134 | | |