MPFRUtils.cpp
19.8 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
//===-- Utils which wrap MPFR ---------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include "MPFRUtils.h"
#include "utils/FPUtil/FPBits.h"
#include "utils/FPUtil/TestHelpers.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/ADT/StringRef.h"
#include <memory>
#include <stdint.h>
#include <string>
#ifdef CUSTOM_MPFR_INCLUDER
// Some downstream repos are monoliths carrying MPFR sources in their third
// party directory. In such repos, including the MPFR header as
// `#include <mpfr.h>` is either disallowed or not possible. If that is the
// case, a file named `CustomMPFRIncluder.h` should be added through which the
// MPFR header can be included in manner allowed in that repo.
#include "CustomMPFRIncluder.h"
#else
#include <mpfr.h>
#endif
template <typename T> using FPBits = __llvm_libc::fputil::FPBits<T>;
namespace __llvm_libc {
namespace testing {
namespace mpfr {
class MPFRNumber {
// A precision value which allows sufficiently large additional
// precision even compared to quad-precision floating point values.
static constexpr unsigned int mpfrPrecision = 128;
mpfr_t value;
public:
MPFRNumber() { mpfr_init2(value, mpfrPrecision); }
// We use explicit EnableIf specializations to disallow implicit
// conversions. Implicit conversions can potentially lead to loss of
// precision.
template <typename XType,
cpp::EnableIfType<cpp::IsSame<float, XType>::Value, int> = 0>
explicit MPFRNumber(XType x) {
mpfr_init2(value, mpfrPrecision);
mpfr_set_flt(value, x, MPFR_RNDN);
}
template <typename XType,
cpp::EnableIfType<cpp::IsSame<double, XType>::Value, int> = 0>
explicit MPFRNumber(XType x) {
mpfr_init2(value, mpfrPrecision);
mpfr_set_d(value, x, MPFR_RNDN);
}
template <typename XType,
cpp::EnableIfType<cpp::IsSame<long double, XType>::Value, int> = 0>
explicit MPFRNumber(XType x) {
mpfr_init2(value, mpfrPrecision);
mpfr_set_ld(value, x, MPFR_RNDN);
}
template <typename XType,
cpp::EnableIfType<cpp::IsIntegral<XType>::Value, int> = 0>
explicit MPFRNumber(XType x) {
mpfr_init2(value, mpfrPrecision);
mpfr_set_sj(value, x, MPFR_RNDN);
}
MPFRNumber(const MPFRNumber &other) {
mpfr_set(value, other.value, MPFR_RNDN);
}
~MPFRNumber() {
mpfr_clear(value);
}
MPFRNumber &operator=(const MPFRNumber &rhs) {
mpfr_set(value, rhs.value, MPFR_RNDN);
return *this;
}
MPFRNumber abs() const {
MPFRNumber result;
mpfr_abs(result.value, value, MPFR_RNDN);
return result;
}
MPFRNumber ceil() const {
MPFRNumber result;
mpfr_ceil(result.value, value);
return result;
}
MPFRNumber cos() const {
MPFRNumber result;
mpfr_cos(result.value, value, MPFR_RNDN);
return result;
}
MPFRNumber exp() const {
MPFRNumber result;
mpfr_exp(result.value, value, MPFR_RNDN);
return result;
}
MPFRNumber exp2() const {
MPFRNumber result;
mpfr_exp2(result.value, value, MPFR_RNDN);
return result;
}
MPFRNumber floor() const {
MPFRNumber result;
mpfr_floor(result.value, value);
return result;
}
MPFRNumber frexp(int &exp) {
MPFRNumber result;
mpfr_exp_t resultExp;
mpfr_frexp(&resultExp, result.value, value, MPFR_RNDN);
exp = resultExp;
return result;
}
MPFRNumber hypot(const MPFRNumber &b) {
MPFRNumber result;
mpfr_hypot(result.value, value, b.value, MPFR_RNDN);
return result;
}
MPFRNumber remquo(const MPFRNumber &divisor, int "ient) {
MPFRNumber remainder;
long q;
mpfr_remquo(remainder.value, &q, value, divisor.value, MPFR_RNDN);
quotient = q;
return remainder;
}
MPFRNumber round() const {
MPFRNumber result;
mpfr_round(result.value, value);
return result;
}
MPFRNumber sin() const {
MPFRNumber result;
mpfr_sin(result.value, value, MPFR_RNDN);
return result;
}
MPFRNumber sqrt() const {
MPFRNumber result;
mpfr_sqrt(result.value, value, MPFR_RNDN);
return result;
}
MPFRNumber trunc() const {
MPFRNumber result;
mpfr_trunc(result.value, value);
return result;
}
std::string str() const {
// 200 bytes should be more than sufficient to hold a 100-digit number
// plus additional bytes for the decimal point, '-' sign etc.
constexpr size_t printBufSize = 200;
char buffer[printBufSize];
mpfr_snprintf(buffer, printBufSize, "%100.50Rf", value);
llvm::StringRef ref(buffer);
ref = ref.trim();
return ref.str();
}
// These functions are useful for debugging.
template <typename T> T as() const;
template <> float as<float>() const { return mpfr_get_flt(value, MPFR_RNDN); }
template <> double as<double>() const { return mpfr_get_d(value, MPFR_RNDN); }
template <> long double as<long double>() const {
return mpfr_get_ld(value, MPFR_RNDN);
}
void dump(const char *msg) const { mpfr_printf("%s%.128Rf\n", msg, value); }
// Return the ULP (units-in-the-last-place) difference between the
// stored MPFR and a floating point number.
//
// We define:
// ULP(mpfr_value, value) = abs(mpfr_value - value) / eps(value)
//
// Remarks:
// 1. ULP < 0.5 will imply that the value is correctly rounded.
// 2. We expect that this value and the value to be compared (the [input]
// argument) are reasonable close, and we will provide an upper bound
// of ULP value for testing. Morever, most of the fractional parts of
// ULP value do not matter much, so using double as the return type
// should be good enough.
template <typename T>
cpp::EnableIfType<cpp::IsFloatingPointType<T>::Value, double> ulp(T input) {
fputil::FPBits<T> bits(input);
MPFRNumber mpfrInput(input);
// abs(value - input)
mpfr_sub(mpfrInput.value, value, mpfrInput.value, MPFR_RNDN);
mpfr_abs(mpfrInput.value, mpfrInput.value, MPFR_RNDN);
// get eps(input)
int epsExponent = bits.exponent - fputil::FPBits<T>::exponentBias -
fputil::MantissaWidth<T>::value;
if (bits.exponent == 0) {
// correcting denormal exponent
++epsExponent;
} else if ((bits.mantissa == 0) && (bits.exponent > 1) &&
mpfr_less_p(value, mpfrInput.value)) {
// when the input is exactly 2^n, distance (epsilon) between the input
// and the next floating point number is different from the distance to
// the previous floating point number. So in that case, if the correct
// value from MPFR is smaller than the input, we use the smaller epsilon
--epsExponent;
}
// Since eps(value) is of the form 2^e, instead of dividing such number,
// we multiply by its inverse 2^{-e}.
mpfr_mul_2si(mpfrInput.value, mpfrInput.value, -epsExponent, MPFR_RNDN);
return mpfrInput.as<double>();
}
};
namespace internal {
template <typename InputType>
cpp::EnableIfType<cpp::IsFloatingPointType<InputType>::Value, MPFRNumber>
unaryOperation(Operation op, InputType input) {
MPFRNumber mpfrInput(input);
switch (op) {
case Operation::Abs:
return mpfrInput.abs();
case Operation::Ceil:
return mpfrInput.ceil();
case Operation::Cos:
return mpfrInput.cos();
case Operation::Exp:
return mpfrInput.exp();
case Operation::Exp2:
return mpfrInput.exp2();
case Operation::Floor:
return mpfrInput.floor();
case Operation::Round:
return mpfrInput.round();
case Operation::Sin:
return mpfrInput.sin();
case Operation::Sqrt:
return mpfrInput.sqrt();
case Operation::Trunc:
return mpfrInput.trunc();
default:
__builtin_unreachable();
}
}
template <typename InputType>
cpp::EnableIfType<cpp::IsFloatingPointType<InputType>::Value, MPFRNumber>
unaryOperationTwoOutputs(Operation op, InputType input, int &output) {
MPFRNumber mpfrInput(input);
switch (op) {
case Operation::Frexp:
return mpfrInput.frexp(output);
default:
__builtin_unreachable();
}
}
template <typename InputType>
cpp::EnableIfType<cpp::IsFloatingPointType<InputType>::Value, MPFRNumber>
binaryOperationOneOutput(Operation op, InputType x, InputType y) {
MPFRNumber inputX(x), inputY(y);
switch (op) {
case Operation::Hypot:
return inputX.hypot(inputY);
default:
__builtin_unreachable();
}
}
template <typename InputType>
cpp::EnableIfType<cpp::IsFloatingPointType<InputType>::Value, MPFRNumber>
binaryOperationTwoOutputs(Operation op, InputType x, InputType y, int &output) {
MPFRNumber inputX(x), inputY(y);
switch (op) {
case Operation::RemQuo:
return inputX.remquo(inputY, output);
default:
__builtin_unreachable();
}
}
template <typename T>
void explainUnaryOperationSingleOutputError(Operation op, T input, T matchValue,
testutils::StreamWrapper &OS) {
MPFRNumber mpfrInput(input);
MPFRNumber mpfrResult = unaryOperation(op, input);
MPFRNumber mpfrMatchValue(matchValue);
FPBits<T> inputBits(input);
FPBits<T> matchBits(matchValue);
FPBits<T> mpfrResultBits(mpfrResult.as<T>());
OS << "Match value not within tolerance value of MPFR result:\n"
<< " Input decimal: " << mpfrInput.str() << '\n';
__llvm_libc::fputil::testing::describeValue(" Input bits: ", input, OS);
OS << '\n' << " Match decimal: " << mpfrMatchValue.str() << '\n';
__llvm_libc::fputil::testing::describeValue(" Match bits: ", matchValue,
OS);
OS << '\n' << " MPFR result: " << mpfrResult.str() << '\n';
__llvm_libc::fputil::testing::describeValue(
" MPFR rounded: ", mpfrResult.as<T>(), OS);
OS << '\n';
OS << " ULP error: " << std::to_string(mpfrResult.ulp(matchValue))
<< '\n';
}
template void
explainUnaryOperationSingleOutputError<float>(Operation op, float, float,
testutils::StreamWrapper &);
template void
explainUnaryOperationSingleOutputError<double>(Operation op, double, double,
testutils::StreamWrapper &);
template void explainUnaryOperationSingleOutputError<long double>(
Operation op, long double, long double, testutils::StreamWrapper &);
template <typename T>
void explainUnaryOperationTwoOutputsError(Operation op, T input,
const BinaryOutput<T> &libcResult,
testutils::StreamWrapper &OS) {
MPFRNumber mpfrInput(input);
FPBits<T> inputBits(input);
int mpfrIntResult;
MPFRNumber mpfrResult = unaryOperationTwoOutputs(op, input, mpfrIntResult);
if (mpfrIntResult != libcResult.i) {
OS << "MPFR integral result: " << mpfrIntResult << '\n'
<< "Libc integral result: " << libcResult.i << '\n';
} else {
OS << "Integral result from libc matches integral result from MPFR.\n";
}
MPFRNumber mpfrMatchValue(libcResult.f);
OS << "Libc floating point result is not within tolerance value of the MPFR "
<< "result.\n\n";
OS << " Input decimal: " << mpfrInput.str() << "\n\n";
OS << "Libc floating point value: " << mpfrMatchValue.str() << '\n';
__llvm_libc::fputil::testing::describeValue(
" Libc floating point bits: ", libcResult.f, OS);
OS << "\n\n";
OS << " MPFR result: " << mpfrResult.str() << '\n';
__llvm_libc::fputil::testing::describeValue(
" MPFR rounded: ", mpfrResult.as<T>(), OS);
OS << '\n'
<< " ULP error: "
<< std::to_string(mpfrResult.ulp(libcResult.f)) << '\n';
}
template void explainUnaryOperationTwoOutputsError<float>(
Operation, float, const BinaryOutput<float> &, testutils::StreamWrapper &);
template void
explainUnaryOperationTwoOutputsError<double>(Operation, double,
const BinaryOutput<double> &,
testutils::StreamWrapper &);
template void explainUnaryOperationTwoOutputsError<long double>(
Operation, long double, const BinaryOutput<long double> &,
testutils::StreamWrapper &);
template <typename T>
void explainBinaryOperationTwoOutputsError(Operation op,
const BinaryInput<T> &input,
const BinaryOutput<T> &libcResult,
testutils::StreamWrapper &OS) {
MPFRNumber mpfrX(input.x);
MPFRNumber mpfrY(input.y);
FPBits<T> xbits(input.x);
FPBits<T> ybits(input.y);
int mpfrIntResult;
MPFRNumber mpfrResult =
binaryOperationTwoOutputs(op, input.x, input.y, mpfrIntResult);
MPFRNumber mpfrMatchValue(libcResult.f);
OS << "Input decimal: x: " << mpfrX.str() << " y: " << mpfrY.str() << '\n'
<< "MPFR integral result: " << mpfrIntResult << '\n'
<< "Libc integral result: " << libcResult.i << '\n'
<< "Libc floating point result: " << mpfrMatchValue.str() << '\n'
<< " MPFR result: " << mpfrResult.str() << '\n';
__llvm_libc::fputil::testing::describeValue(
"Libc floating point result bits: ", libcResult.f, OS);
__llvm_libc::fputil::testing::describeValue(
" MPFR rounded bits: ", mpfrResult.as<T>(), OS);
OS << "ULP error: " << std::to_string(mpfrResult.ulp(libcResult.f)) << '\n';
}
template void explainBinaryOperationTwoOutputsError<float>(
Operation, const BinaryInput<float> &, const BinaryOutput<float> &,
testutils::StreamWrapper &);
template void explainBinaryOperationTwoOutputsError<double>(
Operation, const BinaryInput<double> &, const BinaryOutput<double> &,
testutils::StreamWrapper &);
template void explainBinaryOperationTwoOutputsError<long double>(
Operation, const BinaryInput<long double> &,
const BinaryOutput<long double> &, testutils::StreamWrapper &);
template <typename T>
void explainBinaryOperationOneOutputError(Operation op,
const BinaryInput<T> &input,
T libcResult,
testutils::StreamWrapper &OS) {
MPFRNumber mpfrX(input.x);
MPFRNumber mpfrY(input.y);
FPBits<T> xbits(input.x);
FPBits<T> ybits(input.y);
MPFRNumber mpfrResult = binaryOperationOneOutput(op, input.x, input.y);
MPFRNumber mpfrMatchValue(libcResult);
OS << "Input decimal: x: " << mpfrX.str() << " y: " << mpfrY.str() << '\n';
__llvm_libc::fputil::testing::describeValue("First input bits: ", input.x,
OS);
__llvm_libc::fputil::testing::describeValue("Second input bits: ", input.y,
OS);
OS << "Libc result: " << mpfrMatchValue.str() << '\n'
<< "MPFR result: " << mpfrResult.str() << '\n';
__llvm_libc::fputil::testing::describeValue(
"Libc floating point result bits: ", libcResult, OS);
__llvm_libc::fputil::testing::describeValue(
" MPFR rounded bits: ", mpfrResult.as<T>(), OS);
OS << "ULP error: " << std::to_string(mpfrResult.ulp(libcResult)) << '\n';
}
template void explainBinaryOperationOneOutputError<float>(
Operation, const BinaryInput<float> &, float, testutils::StreamWrapper &);
template void explainBinaryOperationOneOutputError<double>(
Operation, const BinaryInput<double> &, double, testutils::StreamWrapper &);
template void explainBinaryOperationOneOutputError<long double>(
Operation, const BinaryInput<long double> &, long double,
testutils::StreamWrapper &);
template <typename T>
bool compareUnaryOperationSingleOutput(Operation op, T input, T libcResult,
double ulpError) {
// If the ulp error is exactly 0.5 (i.e a tie), we would check that the result
// is rounded to the nearest even.
MPFRNumber mpfrResult = unaryOperation(op, input);
double ulp = mpfrResult.ulp(libcResult);
bool bitsAreEven = ((FPBits<T>(libcResult).bitsAsUInt() & 1) == 0);
return (ulp < ulpError) ||
((ulp == ulpError) && ((ulp != 0.5) || bitsAreEven));
}
template bool compareUnaryOperationSingleOutput<float>(Operation, float, float,
double);
template bool compareUnaryOperationSingleOutput<double>(Operation, double,
double, double);
template bool compareUnaryOperationSingleOutput<long double>(Operation,
long double,
long double,
double);
template <typename T>
bool compareUnaryOperationTwoOutputs(Operation op, T input,
const BinaryOutput<T> &libcResult,
double ulpError) {
int mpfrIntResult;
MPFRNumber mpfrResult = unaryOperationTwoOutputs(op, input, mpfrIntResult);
double ulp = mpfrResult.ulp(libcResult.f);
if (mpfrIntResult != libcResult.i)
return false;
bool bitsAreEven = ((FPBits<T>(libcResult.f).bitsAsUInt() & 1) == 0);
return (ulp < ulpError) ||
((ulp == ulpError) && ((ulp != 0.5) || bitsAreEven));
}
template bool
compareUnaryOperationTwoOutputs<float>(Operation, float,
const BinaryOutput<float> &, double);
template bool
compareUnaryOperationTwoOutputs<double>(Operation, double,
const BinaryOutput<double> &, double);
template bool compareUnaryOperationTwoOutputs<long double>(
Operation, long double, const BinaryOutput<long double> &, double);
template <typename T>
bool compareBinaryOperationTwoOutputs(Operation op, const BinaryInput<T> &input,
const BinaryOutput<T> &libcResult,
double ulpError) {
int mpfrIntResult;
MPFRNumber mpfrResult =
binaryOperationTwoOutputs(op, input.x, input.y, mpfrIntResult);
double ulp = mpfrResult.ulp(libcResult.f);
if (mpfrIntResult != libcResult.i) {
if (op == Operation::RemQuo) {
if ((0x7 & mpfrIntResult) != (0x7 & libcResult.i))
return false;
} else {
return false;
}
}
bool bitsAreEven = ((FPBits<T>(libcResult.f).bitsAsUInt() & 1) == 0);
return (ulp < ulpError) ||
((ulp == ulpError) && ((ulp != 0.5) || bitsAreEven));
}
template bool
compareBinaryOperationTwoOutputs<float>(Operation, const BinaryInput<float> &,
const BinaryOutput<float> &, double);
template bool
compareBinaryOperationTwoOutputs<double>(Operation, const BinaryInput<double> &,
const BinaryOutput<double> &, double);
template bool compareBinaryOperationTwoOutputs<long double>(
Operation, const BinaryInput<long double> &,
const BinaryOutput<long double> &, double);
template <typename T>
bool compareBinaryOperationOneOutput(Operation op, const BinaryInput<T> &input,
T libcResult, double ulpError) {
MPFRNumber mpfrResult = binaryOperationOneOutput(op, input.x, input.y);
double ulp = mpfrResult.ulp(libcResult);
bool bitsAreEven = ((FPBits<T>(libcResult).bitsAsUInt() & 1) == 0);
return (ulp < ulpError) ||
((ulp == ulpError) && ((ulp != 0.5) || bitsAreEven));
}
template bool compareBinaryOperationOneOutput<float>(Operation,
const BinaryInput<float> &,
float, double);
template bool
compareBinaryOperationOneOutput<double>(Operation, const BinaryInput<double> &,
double, double);
template bool compareBinaryOperationOneOutput<long double>(
Operation, const BinaryInput<long double> &, long double, double);
} // namespace internal
} // namespace mpfr
} // namespace testing
} // namespace __llvm_libc