-
Notifications
You must be signed in to change notification settings - Fork 181
Expand file tree
/
Copy pathfloat_common.h
More file actions
1432 lines (1223 loc) · 44 KB
/
float_common.h
File metadata and controls
1432 lines (1223 loc) · 44 KB
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
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#ifndef FASTFLOAT_FLOAT_COMMON_H
#define FASTFLOAT_FLOAT_COMMON_H
#include <cfloat>
#include <cstddef>
#include <cstdint>
#include <cassert>
#include <cstring>
#include <limits>
#include <type_traits>
#include <system_error>
#ifdef __has_include
#if __has_include(<stdfloat>) && (__cplusplus > 202002L || (defined(_MSVC_LANG) && (_MSVC_LANG > 202002L)))
#include <stdfloat>
#endif
#endif
#include "constexpr_feature_detect.h"
#define FASTFLOAT_VERSION_MAJOR 8
#define FASTFLOAT_VERSION_MINOR 2
#define FASTFLOAT_VERSION_PATCH 5
#define FASTFLOAT_STRINGIZE_IMPL(x) #x
#define FASTFLOAT_STRINGIZE(x) FASTFLOAT_STRINGIZE_IMPL(x)
#define FASTFLOAT_VERSION_STR \
FASTFLOAT_STRINGIZE(FASTFLOAT_VERSION_MAJOR) \
"." FASTFLOAT_STRINGIZE(FASTFLOAT_VERSION_MINOR) "." FASTFLOAT_STRINGIZE( \
FASTFLOAT_VERSION_PATCH)
#define FASTFLOAT_VERSION \
(FASTFLOAT_VERSION_MAJOR * 10000 + FASTFLOAT_VERSION_MINOR * 100 + \
FASTFLOAT_VERSION_PATCH)
namespace fast_float {
enum class chars_format : uint64_t;
namespace detail {
constexpr chars_format basic_json_fmt = chars_format(1 << 5);
constexpr chars_format basic_fortran_fmt = chars_format(1 << 6);
} // namespace detail
enum class chars_format : uint64_t {
scientific = 1 << 0,
fixed = 1 << 2,
hex = 1 << 3,
no_infnan = 1 << 4,
// RFC 8259: https://datatracker.ietf.org/doc/html/rfc8259#section-6
json = uint64_t(detail::basic_json_fmt) | fixed | scientific | no_infnan,
// Extension of RFC 8259 where, e.g., "inf" and "nan" are allowed.
json_or_infnan = uint64_t(detail::basic_json_fmt) | fixed | scientific,
fortran = uint64_t(detail::basic_fortran_fmt) | fixed | scientific,
general = fixed | scientific,
allow_leading_plus = 1 << 7,
skip_white_space = 1 << 8,
};
template <typename UC> struct from_chars_result_t {
UC const *ptr;
std::errc ec;
// https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/p2497r0.html
constexpr explicit operator bool() const noexcept {
return ec == std::errc();
}
};
using from_chars_result = from_chars_result_t<char>;
template <typename UC> struct parse_options_t {
constexpr explicit parse_options_t(chars_format fmt = chars_format::general,
UC dot = UC('.'), int b = 10)
: format(fmt), decimal_point(dot), base(b) {}
/** Which number formats are accepted */
chars_format format;
/** The character used as decimal point */
UC decimal_point;
/** The base used for integers */
int base;
};
using parse_options = parse_options_t<char>;
} // namespace fast_float
#if FASTFLOAT_HAS_BIT_CAST
#include <bit>
#endif
#if (defined(__x86_64) || defined(__x86_64__) || defined(_M_X64) || \
defined(__amd64) || defined(__aarch64__) || defined(_M_ARM64) || \
defined(__MINGW64__) || defined(__s390x__) || \
(defined(__ppc64__) || defined(__PPC64__) || defined(__ppc64le__) || \
defined(__PPC64LE__)) || \
defined(__loongarch64) || (defined(__riscv) && __riscv_xlen == 64))
#define FASTFLOAT_64BIT 1
#elif (defined(__i386) || defined(__i386__) || defined(_M_IX86) || \
defined(__arm__) || defined(_M_ARM) || defined(__ppc__) || \
defined(__MINGW32__) || defined(__EMSCRIPTEN__) || \
(defined(__riscv) && __riscv_xlen == 32))
#define FASTFLOAT_32BIT 1
#else
// Need to check incrementally, since SIZE_MAX is a size_t, avoid overflow.
// We can never tell the register width, but the SIZE_MAX is a good
// approximation. UINTPTR_MAX and INTPTR_MAX are optional, so avoid them for max
// portability.
#if SIZE_MAX == 0xffff
#error Unknown platform (16-bit, unsupported)
#elif SIZE_MAX == 0xffffffff
#define FASTFLOAT_32BIT 1
#elif SIZE_MAX == 0xffffffffffffffff
#define FASTFLOAT_64BIT 1
#else
#error Unknown platform (not 32-bit, not 64-bit?)
#endif
#endif
#if ((defined(_WIN32) || defined(_WIN64)) && !defined(__clang__)) || \
(defined(_M_ARM64) && !defined(__MINGW32__))
#include <intrin.h>
#endif
#if defined(_MSC_VER) && !defined(__clang__)
#define FASTFLOAT_VISUAL_STUDIO 1
#endif
#if defined __BYTE_ORDER__ && defined __ORDER_BIG_ENDIAN__
#define FASTFLOAT_IS_BIG_ENDIAN (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
#elif defined _WIN32
#define FASTFLOAT_IS_BIG_ENDIAN 0
#else
#if defined(__APPLE__) || defined(__FreeBSD__)
#include <machine/endian.h>
#elif defined(sun) || defined(__sun)
#include <sys/byteorder.h>
#elif defined(__MVS__)
#include <sys/endian.h>
#else
#ifdef __has_include
#if __has_include(<endian.h>)
#include <endian.h>
#endif //__has_include(<endian.h>)
#endif //__has_include
#endif
#
#ifndef __BYTE_ORDER__
// safe choice
#define FASTFLOAT_IS_BIG_ENDIAN 0
#endif
#
#ifndef __ORDER_LITTLE_ENDIAN__
// safe choice
#define FASTFLOAT_IS_BIG_ENDIAN 0
#endif
#
#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
#define FASTFLOAT_IS_BIG_ENDIAN 0
#else
#define FASTFLOAT_IS_BIG_ENDIAN 1
#endif
#endif
#if defined(__SSE2__) || (defined(FASTFLOAT_VISUAL_STUDIO) && \
(defined(_M_AMD64) || defined(_M_X64) || \
(defined(_M_IX86_FP) && _M_IX86_FP == 2)))
#define FASTFLOAT_SSE2 1
#endif
#if defined(__aarch64__) || defined(_M_ARM64)
#define FASTFLOAT_NEON 1
#endif
#if defined(FASTFLOAT_SSE2) || defined(FASTFLOAT_NEON)
#define FASTFLOAT_HAS_SIMD 1
#endif
#if defined(__GNUC__)
// disable -Wcast-align=strict (GCC only)
#define FASTFLOAT_SIMD_DISABLE_WARNINGS \
_Pragma("GCC diagnostic push") \
_Pragma("GCC diagnostic ignored \"-Wcast-align\"")
#else
#define FASTFLOAT_SIMD_DISABLE_WARNINGS
#endif
#if defined(__GNUC__)
#define FASTFLOAT_SIMD_RESTORE_WARNINGS _Pragma("GCC diagnostic pop")
#else
#define FASTFLOAT_SIMD_RESTORE_WARNINGS
#endif
#ifdef FASTFLOAT_VISUAL_STUDIO
#define fastfloat_really_inline __forceinline
#else
#define fastfloat_really_inline inline __attribute__((always_inline))
#endif
#ifndef FASTFLOAT_ASSERT
#define FASTFLOAT_ASSERT(x) \
{ ((void)(x)); }
#endif
#ifndef FASTFLOAT_DEBUG_ASSERT
#define FASTFLOAT_DEBUG_ASSERT(x) \
{ ((void)(x)); }
#endif
// rust style `try!()` macro, or `?` operator
#define FASTFLOAT_TRY(x) \
{ \
if (!(x)) \
return false; \
}
#define FASTFLOAT_ENABLE_IF(...) \
typename std::enable_if<(__VA_ARGS__), int>::type
namespace fast_float {
fastfloat_really_inline constexpr bool cpp20_and_in_constexpr() {
#if FASTFLOAT_HAS_IS_CONSTANT_EVALUATED
return std::is_constant_evaluated();
#else
return false;
#endif
}
template <typename T>
struct is_supported_float_type
: std::integral_constant<
bool, std::is_same<T, double>::value || std::is_same<T, float>::value
#ifdef __STDCPP_FLOAT64_T__
|| std::is_same<T, std::float64_t>::value
#endif
#ifdef __STDCPP_FLOAT32_T__
|| std::is_same<T, std::float32_t>::value
#endif
#ifdef __STDCPP_FLOAT16_T__
|| std::is_same<T, std::float16_t>::value
#endif
#ifdef __STDCPP_BFLOAT16_T__
|| std::is_same<T, std::bfloat16_t>::value
#endif
> {
};
template <typename T>
using equiv_uint_t = typename std::conditional<
sizeof(T) == 1, uint8_t,
typename std::conditional<
sizeof(T) == 2, uint16_t,
typename std::conditional<sizeof(T) == 4, uint32_t,
uint64_t>::type>::type>::type;
template <typename T> struct is_supported_integer_type : std::is_integral<T> {};
template <typename UC>
struct is_supported_char_type
: std::integral_constant<bool, std::is_same<UC, char>::value ||
std::is_same<UC, wchar_t>::value ||
std::is_same<UC, char16_t>::value ||
std::is_same<UC, char32_t>::value
#ifdef __cpp_char8_t
|| std::is_same<UC, char8_t>::value
#endif
> {
};
template <typename UC>
inline FASTFLOAT_CONSTEXPR14 bool
fastfloat_strncasecmp3(UC const *actual_mixedcase,
UC const *expected_lowercase) {
uint64_t mask{0};
FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 1) { mask = 0x2020202020202020; }
else FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 2) {
mask = 0x0020002000200020;
}
else FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 4) {
mask = 0x0000002000000020;
}
else {
return false;
}
uint64_t val1{0}, val2{0};
if (cpp20_and_in_constexpr()) {
for (size_t i = 0; i < 3; i++) {
if ((actual_mixedcase[i] | 32) != expected_lowercase[i]) {
return false;
}
}
return true;
} else {
FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 1 || sizeof(UC) == 2) {
::memcpy(&val1, actual_mixedcase, 3 * sizeof(UC));
::memcpy(&val2, expected_lowercase, 3 * sizeof(UC));
val1 |= mask;
val2 |= mask;
return val1 == val2;
}
else FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 4) {
::memcpy(&val1, actual_mixedcase, 2 * sizeof(UC));
::memcpy(&val2, expected_lowercase, 2 * sizeof(UC));
val1 |= mask;
if (val1 != val2) {
return false;
}
return (actual_mixedcase[2] | 32) == (expected_lowercase[2]);
}
else {
return false;
}
}
}
template <typename UC>
inline FASTFLOAT_CONSTEXPR14 bool
fastfloat_strncasecmp5(UC const *actual_mixedcase,
UC const *expected_lowercase) {
uint64_t mask{0};
uint64_t val1{0}, val2{0};
if (cpp20_and_in_constexpr()) {
for (size_t i = 0; i < 5; i++) {
if ((actual_mixedcase[i] | 32) != expected_lowercase[i]) {
return false;
}
}
return true;
} else {
FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 1) {
mask = 0x2020202020202020;
::memcpy(&val1, actual_mixedcase, 5 * sizeof(UC));
::memcpy(&val2, expected_lowercase, 5 * sizeof(UC));
val1 |= mask;
val2 |= mask;
return val1 == val2;
}
else FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 2) {
mask = 0x0020002000200020;
::memcpy(&val1, actual_mixedcase, 4 * sizeof(UC));
::memcpy(&val2, expected_lowercase, 4 * sizeof(UC));
val1 |= mask;
if (val1 != val2) {
return false;
}
return (actual_mixedcase[4] | 32) == (expected_lowercase[4]);
}
else FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 4) {
mask = 0x0000002000000020;
::memcpy(&val1, actual_mixedcase, 2 * sizeof(UC));
::memcpy(&val2, expected_lowercase, 2 * sizeof(UC));
val1 |= mask;
if (val1 != val2) {
return false;
}
::memcpy(&val1, actual_mixedcase + 2, 2 * sizeof(UC));
::memcpy(&val2, expected_lowercase + 2, 2 * sizeof(UC));
val1 |= mask;
if (val1 != val2) {
return false;
}
return (actual_mixedcase[4] | 32) == (expected_lowercase[4]);
}
else {
return false;
}
}
}
// Compares two ASCII strings in a case insensitive manner.
template <typename UC>
inline FASTFLOAT_CONSTEXPR14 bool
fastfloat_strncasecmp(UC const *actual_mixedcase, UC const *expected_lowercase,
size_t length) {
uint64_t mask{0};
FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 1) { mask = 0x2020202020202020; }
else FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 2) {
mask = 0x0020002000200020;
}
else FASTFLOAT_IF_CONSTEXPR17(sizeof(UC) == 4) {
mask = 0x0000002000000020;
}
else {
return false;
}
if (cpp20_and_in_constexpr()) {
for (size_t i = 0; i < length; i++) {
if ((actual_mixedcase[i] | 32) != expected_lowercase[i]) {
return false;
}
}
return true;
} else {
uint64_t val1{0}, val2{0};
size_t sz{8 / (sizeof(UC))};
for (size_t i = 0; i < length; i += sz) {
val1 = val2 = 0;
sz = sz < (length - i) ? sz : length - i;
::memcpy(&val1, actual_mixedcase + i, sz * sizeof(UC));
::memcpy(&val2, expected_lowercase + i, sz * sizeof(UC));
val1 |= mask;
val2 |= mask;
if (val1 != val2) {
return false;
}
}
return true;
}
}
#ifndef FLT_EVAL_METHOD
#error "FLT_EVAL_METHOD should be defined, please include cfloat."
#endif
// a pointer and a length to a contiguous block of memory
template <typename T> struct span {
T const *ptr;
size_t length;
constexpr span(T const *_ptr, size_t _length) : ptr(_ptr), length(_length) {}
constexpr span() : ptr(nullptr), length(0) {}
constexpr size_t len() const noexcept { return length; }
FASTFLOAT_CONSTEXPR14 const T &operator[](size_t index) const noexcept {
FASTFLOAT_DEBUG_ASSERT(index < length);
return ptr[index];
}
};
struct value128 {
uint64_t low;
uint64_t high;
constexpr value128(uint64_t _low, uint64_t _high) : low(_low), high(_high) {}
constexpr value128() : low(0), high(0) {}
};
/* Helper C++14 constexpr generic implementation of leading_zeroes */
fastfloat_really_inline FASTFLOAT_CONSTEXPR14 int
leading_zeroes_generic(uint64_t input_num, int last_bit = 0) {
if (input_num & uint64_t(0xffffffff00000000)) {
input_num >>= 32;
last_bit |= 32;
}
if (input_num & uint64_t(0xffff0000)) {
input_num >>= 16;
last_bit |= 16;
}
if (input_num & uint64_t(0xff00)) {
input_num >>= 8;
last_bit |= 8;
}
if (input_num & uint64_t(0xf0)) {
input_num >>= 4;
last_bit |= 4;
}
if (input_num & uint64_t(0xc)) {
input_num >>= 2;
last_bit |= 2;
}
if (input_num & uint64_t(0x2)) { /* input_num >>= 1; */
last_bit |= 1;
}
return 63 - last_bit;
}
/* result might be undefined when input_num is zero */
fastfloat_really_inline FASTFLOAT_CONSTEXPR20 int
leading_zeroes(uint64_t input_num) {
assert(input_num > 0);
if (cpp20_and_in_constexpr()) {
return leading_zeroes_generic(input_num);
}
#ifdef FASTFLOAT_VISUAL_STUDIO
#if defined(_M_X64) || defined(_M_ARM64)
unsigned long leading_zero = 0;
// Search the mask data from most significant bit (MSB)
// to least significant bit (LSB) for a set bit (1).
_BitScanReverse64(&leading_zero, input_num);
return (int)(63 - leading_zero);
#else
return leading_zeroes_generic(input_num);
#endif
#else
return __builtin_clzll(input_num);
#endif
}
/* Helper C++14 constexpr generic implementation of countr_zero for 32-bit */
fastfloat_really_inline FASTFLOAT_CONSTEXPR14 int
countr_zero_generic_32(uint32_t input_num) {
if (input_num == 0) {
return 32;
}
int last_bit = 0;
if (!(input_num & 0x0000FFFF)) {
input_num >>= 16;
last_bit |= 16;
}
if (!(input_num & 0x00FF)) {
input_num >>= 8;
last_bit |= 8;
}
if (!(input_num & 0x0F)) {
input_num >>= 4;
last_bit |= 4;
}
if (!(input_num & 0x3)) {
input_num >>= 2;
last_bit |= 2;
}
if (!(input_num & 0x1)) {
last_bit |= 1;
}
return last_bit;
}
/* count trailing zeroes for 32-bit integers */
fastfloat_really_inline FASTFLOAT_CONSTEXPR20 int
countr_zero_32(uint32_t input_num) {
if (cpp20_and_in_constexpr()) {
return countr_zero_generic_32(input_num);
}
#ifdef FASTFLOAT_VISUAL_STUDIO
unsigned long trailing_zero = 0;
if (_BitScanForward(&trailing_zero, input_num)) {
return (int)trailing_zero;
}
return 32;
#else
return input_num == 0 ? 32 : __builtin_ctz(input_num);
#endif
}
// slow emulation routine for 32-bit
fastfloat_really_inline constexpr uint64_t emulu(uint32_t x, uint32_t y) {
return x * (uint64_t)y;
}
fastfloat_really_inline FASTFLOAT_CONSTEXPR14 uint64_t
umul128_generic(uint64_t ab, uint64_t cd, uint64_t *hi) {
uint64_t ad = emulu((uint32_t)(ab >> 32), (uint32_t)cd);
uint64_t bd = emulu((uint32_t)ab, (uint32_t)cd);
uint64_t adbc = ad + emulu((uint32_t)ab, (uint32_t)(cd >> 32));
uint64_t adbc_carry = (uint64_t)(adbc < ad);
uint64_t lo = bd + (adbc << 32);
*hi = emulu((uint32_t)(ab >> 32), (uint32_t)(cd >> 32)) + (adbc >> 32) +
(adbc_carry << 32) + (uint64_t)(lo < bd);
return lo;
}
#ifdef FASTFLOAT_32BIT
// slow emulation routine for 32-bit
#if !defined(__MINGW64__)
fastfloat_really_inline FASTFLOAT_CONSTEXPR14 uint64_t _umul128(uint64_t ab,
uint64_t cd,
uint64_t *hi) {
return umul128_generic(ab, cd, hi);
}
#endif // !__MINGW64__
#endif // FASTFLOAT_32BIT
// compute 64-bit a*b
fastfloat_really_inline FASTFLOAT_CONSTEXPR20 value128
full_multiplication(uint64_t a, uint64_t b) {
if (cpp20_and_in_constexpr()) {
value128 answer;
answer.low = umul128_generic(a, b, &answer.high);
return answer;
}
value128 answer;
#if defined(_M_ARM64) && !defined(__MINGW32__)
// ARM64 has native support for 64-bit multiplications, no need to emulate
// But MinGW on ARM64 doesn't have native support for 64-bit multiplications
answer.high = __umulh(a, b);
answer.low = a * b;
#elif defined(FASTFLOAT_32BIT) || (defined(_WIN64) && !defined(__clang__) && \
!defined(_M_ARM64) && !defined(__GNUC__))
answer.low = _umul128(a, b, &answer.high); // _umul128 not available on ARM64
#elif defined(FASTFLOAT_64BIT) && defined(__SIZEOF_INT128__)
__uint128_t r = ((__uint128_t)a) * b;
answer.low = uint64_t(r);
answer.high = uint64_t(r >> 64);
#else
answer.low = umul128_generic(a, b, &answer.high);
#endif
return answer;
}
struct adjusted_mantissa {
uint64_t mantissa{0};
int32_t power2{0}; // a negative value indicates an invalid result
adjusted_mantissa() = default;
constexpr bool operator==(adjusted_mantissa const &o) const {
return mantissa == o.mantissa && power2 == o.power2;
}
constexpr bool operator!=(adjusted_mantissa const &o) const {
return mantissa != o.mantissa || power2 != o.power2;
}
};
// Bias so we can get the real exponent with an invalid adjusted_mantissa.
constexpr static int32_t invalid_am_bias = -0x8000;
// used for binary_format_lookup_tables<T>::max_mantissa
constexpr uint64_t constant_55555 = 5 * 5 * 5 * 5 * 5;
template <typename T, typename U = void> struct binary_format_lookup_tables;
template <typename T> struct binary_format : binary_format_lookup_tables<T> {
using equiv_uint = equiv_uint_t<T>;
static constexpr int mantissa_explicit_bits();
static constexpr int minimum_exponent();
static constexpr int infinite_power();
static constexpr int sign_index();
static constexpr int
min_exponent_fast_path(); // used when fegetround() == FE_TONEAREST
static constexpr int max_exponent_fast_path();
static constexpr int max_exponent_round_to_even();
static constexpr int min_exponent_round_to_even();
static constexpr uint64_t max_mantissa_fast_path(int64_t power);
static constexpr uint64_t
max_mantissa_fast_path(); // used when fegetround() == FE_TONEAREST
static constexpr int largest_power_of_ten();
static constexpr int smallest_power_of_ten();
static constexpr T exact_power_of_ten(int64_t power);
static constexpr size_t max_digits();
static constexpr equiv_uint exponent_mask();
static constexpr equiv_uint mantissa_mask();
static constexpr equiv_uint hidden_bit_mask();
};
template <typename U> struct binary_format_lookup_tables<double, U> {
static constexpr double powers_of_ten[] = {
1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9, 1e10, 1e11,
1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19, 1e20, 1e21, 1e22};
// Largest integer value v so that (5**index * v) <= 1<<53.
// 0x20000000000000 == 1 << 53
static constexpr uint64_t max_mantissa[] = {
0x20000000000000,
0x20000000000000 / 5,
0x20000000000000 / (5 * 5),
0x20000000000000 / (5 * 5 * 5),
0x20000000000000 / (5 * 5 * 5 * 5),
0x20000000000000 / (constant_55555),
0x20000000000000 / (constant_55555 * 5),
0x20000000000000 / (constant_55555 * 5 * 5),
0x20000000000000 / (constant_55555 * 5 * 5 * 5),
0x20000000000000 / (constant_55555 * 5 * 5 * 5 * 5),
0x20000000000000 / (constant_55555 * constant_55555),
0x20000000000000 / (constant_55555 * constant_55555 * 5),
0x20000000000000 / (constant_55555 * constant_55555 * 5 * 5),
0x20000000000000 / (constant_55555 * constant_55555 * 5 * 5 * 5),
0x20000000000000 / (constant_55555 * constant_55555 * constant_55555),
0x20000000000000 / (constant_55555 * constant_55555 * constant_55555 * 5),
0x20000000000000 /
(constant_55555 * constant_55555 * constant_55555 * 5 * 5),
0x20000000000000 /
(constant_55555 * constant_55555 * constant_55555 * 5 * 5 * 5),
0x20000000000000 /
(constant_55555 * constant_55555 * constant_55555 * 5 * 5 * 5 * 5),
0x20000000000000 /
(constant_55555 * constant_55555 * constant_55555 * constant_55555),
0x20000000000000 / (constant_55555 * constant_55555 * constant_55555 *
constant_55555 * 5),
0x20000000000000 / (constant_55555 * constant_55555 * constant_55555 *
constant_55555 * 5 * 5),
0x20000000000000 / (constant_55555 * constant_55555 * constant_55555 *
constant_55555 * 5 * 5 * 5),
0x20000000000000 / (constant_55555 * constant_55555 * constant_55555 *
constant_55555 * 5 * 5 * 5 * 5)};
};
#if FASTFLOAT_DETAIL_MUST_DEFINE_CONSTEXPR_VARIABLE
template <typename U>
constexpr double binary_format_lookup_tables<double, U>::powers_of_ten[];
template <typename U>
constexpr uint64_t binary_format_lookup_tables<double, U>::max_mantissa[];
#endif
template <typename U> struct binary_format_lookup_tables<float, U> {
static constexpr float powers_of_ten[] = {1e0f, 1e1f, 1e2f, 1e3f, 1e4f, 1e5f,
1e6f, 1e7f, 1e8f, 1e9f, 1e10f};
// Largest integer value v so that (5**index * v) <= 1<<24.
// 0x1000000 == 1<<24
static constexpr uint64_t max_mantissa[] = {
0x1000000,
0x1000000 / 5,
0x1000000 / (5 * 5),
0x1000000 / (5 * 5 * 5),
0x1000000 / (5 * 5 * 5 * 5),
0x1000000 / (constant_55555),
0x1000000 / (constant_55555 * 5),
0x1000000 / (constant_55555 * 5 * 5),
0x1000000 / (constant_55555 * 5 * 5 * 5),
0x1000000 / (constant_55555 * 5 * 5 * 5 * 5),
0x1000000 / (constant_55555 * constant_55555),
0x1000000 / (constant_55555 * constant_55555 * 5)};
};
#if FASTFLOAT_DETAIL_MUST_DEFINE_CONSTEXPR_VARIABLE
template <typename U>
constexpr float binary_format_lookup_tables<float, U>::powers_of_ten[];
template <typename U>
constexpr uint64_t binary_format_lookup_tables<float, U>::max_mantissa[];
#endif
template <>
inline constexpr int binary_format<double>::min_exponent_fast_path() {
#if (FLT_EVAL_METHOD != 1) && (FLT_EVAL_METHOD != 0)
return 0;
#else
return -22;
#endif
}
template <>
inline constexpr int binary_format<float>::min_exponent_fast_path() {
#if (FLT_EVAL_METHOD != 1) && (FLT_EVAL_METHOD != 0)
return 0;
#else
return -10;
#endif
}
template <>
inline constexpr int binary_format<double>::mantissa_explicit_bits() {
return 52;
}
template <>
inline constexpr int binary_format<float>::mantissa_explicit_bits() {
return 23;
}
template <>
inline constexpr int binary_format<double>::max_exponent_round_to_even() {
return 23;
}
template <>
inline constexpr int binary_format<float>::max_exponent_round_to_even() {
return 10;
}
template <>
inline constexpr int binary_format<double>::min_exponent_round_to_even() {
return -4;
}
template <>
inline constexpr int binary_format<float>::min_exponent_round_to_even() {
return -17;
}
template <> inline constexpr int binary_format<double>::minimum_exponent() {
return -1023;
}
template <> inline constexpr int binary_format<float>::minimum_exponent() {
return -127;
}
template <> inline constexpr int binary_format<double>::infinite_power() {
return 0x7FF;
}
template <> inline constexpr int binary_format<float>::infinite_power() {
return 0xFF;
}
template <> inline constexpr int binary_format<double>::sign_index() {
return 63;
}
template <> inline constexpr int binary_format<float>::sign_index() {
return 31;
}
template <>
inline constexpr int binary_format<double>::max_exponent_fast_path() {
return 22;
}
template <>
inline constexpr int binary_format<float>::max_exponent_fast_path() {
return 10;
}
template <>
inline constexpr uint64_t binary_format<double>::max_mantissa_fast_path() {
return uint64_t(2) << mantissa_explicit_bits();
}
template <>
inline constexpr uint64_t binary_format<float>::max_mantissa_fast_path() {
return uint64_t(2) << mantissa_explicit_bits();
}
// credit: Jakub Jelínek
#ifdef __STDCPP_FLOAT16_T__
template <typename U> struct binary_format_lookup_tables<std::float16_t, U> {
static constexpr std::float16_t powers_of_ten[] = {1e0f16, 1e1f16, 1e2f16,
1e3f16, 1e4f16};
// Largest integer value v so that (5**index * v) <= 1<<11.
// 0x800 == 1<<11
static constexpr uint64_t max_mantissa[] = {0x800,
0x800 / 5,
0x800 / (5 * 5),
0x800 / (5 * 5 * 5),
0x800 / (5 * 5 * 5 * 5),
0x800 / (constant_55555)};
};
#if FASTFLOAT_DETAIL_MUST_DEFINE_CONSTEXPR_VARIABLE
template <typename U>
constexpr std::float16_t
binary_format_lookup_tables<std::float16_t, U>::powers_of_ten[];
template <typename U>
constexpr uint64_t
binary_format_lookup_tables<std::float16_t, U>::max_mantissa[];
#endif
template <>
inline constexpr std::float16_t
binary_format<std::float16_t>::exact_power_of_ten(int64_t power) {
// Work around clang bug https://godbolt.org/z/zedh7rrhc
return (void)powers_of_ten[0], powers_of_ten[power];
}
template <>
inline constexpr binary_format<std::float16_t>::equiv_uint
binary_format<std::float16_t>::exponent_mask() {
return 0x7C00;
}
template <>
inline constexpr binary_format<std::float16_t>::equiv_uint
binary_format<std::float16_t>::mantissa_mask() {
return 0x03FF;
}
template <>
inline constexpr binary_format<std::float16_t>::equiv_uint
binary_format<std::float16_t>::hidden_bit_mask() {
return 0x0400;
}
template <>
inline constexpr int binary_format<std::float16_t>::max_exponent_fast_path() {
return 4;
}
template <>
inline constexpr int binary_format<std::float16_t>::mantissa_explicit_bits() {
return 10;
}
template <>
inline constexpr uint64_t
binary_format<std::float16_t>::max_mantissa_fast_path() {
return uint64_t(2) << mantissa_explicit_bits();
}
template <>
inline constexpr uint64_t
binary_format<std::float16_t>::max_mantissa_fast_path(int64_t power) {
// caller is responsible to ensure that
// power >= 0 && power <= 4
//
// Work around clang bug https://godbolt.org/z/zedh7rrhc
return (void)max_mantissa[0], max_mantissa[power];
}
template <>
inline constexpr int binary_format<std::float16_t>::min_exponent_fast_path() {
return 0;
}
template <>
inline constexpr int
binary_format<std::float16_t>::max_exponent_round_to_even() {
return 5;
}
template <>
inline constexpr int
binary_format<std::float16_t>::min_exponent_round_to_even() {
return -22;
}
template <>
inline constexpr int binary_format<std::float16_t>::minimum_exponent() {
return -15;
}
template <>
inline constexpr int binary_format<std::float16_t>::infinite_power() {
return 0x1F;
}
template <> inline constexpr int binary_format<std::float16_t>::sign_index() {
return 15;
}
template <>
inline constexpr int binary_format<std::float16_t>::largest_power_of_ten() {
return 4;
}
template <>
inline constexpr int binary_format<std::float16_t>::smallest_power_of_ten() {
return -27;
}
template <>
inline constexpr size_t binary_format<std::float16_t>::max_digits() {
return 22;
}
#endif // __STDCPP_FLOAT16_T__
// credit: Jakub Jelínek
#ifdef __STDCPP_BFLOAT16_T__
template <typename U> struct binary_format_lookup_tables<std::bfloat16_t, U> {
static constexpr std::bfloat16_t powers_of_ten[] = {1e0bf16, 1e1bf16, 1e2bf16,
1e3bf16};
// Largest integer value v so that (5**index * v) <= 1<<8.
// 0x100 == 1<<8
static constexpr uint64_t max_mantissa[] = {0x100, 0x100 / 5, 0x100 / (5 * 5),
0x100 / (5 * 5 * 5),
0x100 / (5 * 5 * 5 * 5)};
};
#if FASTFLOAT_DETAIL_MUST_DEFINE_CONSTEXPR_VARIABLE
template <typename U>
constexpr std::bfloat16_t
binary_format_lookup_tables<std::bfloat16_t, U>::powers_of_ten[];
template <typename U>
constexpr uint64_t
binary_format_lookup_tables<std::bfloat16_t, U>::max_mantissa[];
#endif
template <>
inline constexpr std::bfloat16_t
binary_format<std::bfloat16_t>::exact_power_of_ten(int64_t power) {
// Work around clang bug https://godbolt.org/z/zedh7rrhc
return (void)powers_of_ten[0], powers_of_ten[power];
}
template <>
inline constexpr int binary_format<std::bfloat16_t>::max_exponent_fast_path() {
return 3;
}
template <>
inline constexpr binary_format<std::bfloat16_t>::equiv_uint
binary_format<std::bfloat16_t>::exponent_mask() {
return 0x7F80;
}
template <>
inline constexpr binary_format<std::bfloat16_t>::equiv_uint
binary_format<std::bfloat16_t>::mantissa_mask() {
return 0x007F;
}
template <>
inline constexpr binary_format<std::bfloat16_t>::equiv_uint
binary_format<std::bfloat16_t>::hidden_bit_mask() {
return 0x0080;
}
template <>