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
|
#include <c10/util/complex.h>
#include <c10/util/math_compat.h>
#include <cmath>
// Note [ Complex Square root in libc++]
// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
// In libc++ complex square root is computed using polar form
// This is a reasonably fast algorithm, but can result in significant
// numerical errors when arg is close to 0, pi/2, pi, or 3pi/4
// In that case provide a more conservative implementation which is
// slower but less prone to those kinds of errors
// In libstdc++ complex square root yield invalid results
// for -x-0.0j unless C99 csqrt/csqrtf fallbacks are used
#if defined(_LIBCPP_VERSION) || \
(defined(__GLIBCXX__) && !defined(_GLIBCXX11_USE_C99_COMPLEX))
namespace {
template <typename T>
c10::complex<T> compute_csqrt(const c10::complex<T>& z) {
constexpr auto half = T(.5);
// Trust standard library to correctly handle infs and NaNs
if (std::isinf(z.real()) || std::isinf(z.imag()) || std::isnan(z.real()) ||
std::isnan(z.imag())) {
return static_cast<c10::complex<T>>(
std::sqrt(static_cast<std::complex<T>>(z)));
}
// Special case for square root of pure imaginary values
if (z.real() == T(0)) {
if (z.imag() == T(0)) {
return c10::complex<T>(T(0), z.imag());
}
auto v = std::sqrt(half * std::abs(z.imag()));
return c10::complex<T>(v, std::copysign(v, z.imag()));
}
// At this point, z is non-zero and finite
if (z.real() >= 0.0) {
auto t = std::sqrt((z.real() + std::abs(z)) * half);
return c10::complex<T>(t, half * (z.imag() / t));
}
auto t = std::sqrt((-z.real() + std::abs(z)) * half);
return c10::complex<T>(
half * std::abs(z.imag() / t), std::copysign(t, z.imag()));
}
// Compute complex arccosine using formula from W. Kahan
// "Branch Cuts for Complex Elementary Functions" 1986 paper:
// cacos(z).re = 2*atan2(sqrt(1-z).re(), sqrt(1+z).re())
// cacos(z).im = asinh((sqrt(conj(1+z))*sqrt(1-z)).im())
template <typename T>
c10::complex<T> compute_cacos(const c10::complex<T>& z) {
auto constexpr one = T(1);
// Trust standard library to correctly handle infs and NaNs
if (std::isinf(z.real()) || std::isinf(z.imag()) || std::isnan(z.real()) ||
std::isnan(z.imag())) {
return static_cast<c10::complex<T>>(
std::acos(static_cast<std::complex<T>>(z)));
}
auto a = compute_csqrt(c10::complex<T>(one - z.real(), -z.imag()));
auto b = compute_csqrt(c10::complex<T>(one + z.real(), z.imag()));
auto c = compute_csqrt(c10::complex<T>(one + z.real(), -z.imag()));
auto r = T(2) * std::atan2(a.real(), b.real());
// Explicitly unroll (a*c).imag()
auto i = std::asinh(a.real() * c.imag() + a.imag() * c.real());
return c10::complex<T>(r, i);
}
} // anonymous namespace
namespace c10_complex_math {
namespace _detail {
c10::complex<float> sqrt(const c10::complex<float>& in) {
return compute_csqrt(in);
}
c10::complex<double> sqrt(const c10::complex<double>& in) {
return compute_csqrt(in);
}
c10::complex<float> acos(const c10::complex<float>& in) {
return compute_cacos(in);
}
c10::complex<double> acos(const c10::complex<double>& in) {
return compute_cacos(in);
}
} // namespace _detail
} // namespace c10_complex_math
#endif
|
