1 | #ifdef __OPENCL_VERSION__ |
---|
2 | # define USE_OPENCL |
---|
3 | #elif defined(_OPENMP) |
---|
4 | # define USE_OPENMP |
---|
5 | #endif |
---|
6 | |
---|
7 | // If opencl is not available, then we are compiling a C function |
---|
8 | // Note: if using a C++ compiler, then define kernel as extern "C" |
---|
9 | #ifndef USE_OPENCL |
---|
10 | # ifdef __cplusplus |
---|
11 | #include <cstdio> |
---|
12 | #include <cmath> |
---|
13 | using namespace std; |
---|
14 | #if defined(_MSC_VER) |
---|
15 | #include <limits> |
---|
16 | #include <float.h> |
---|
17 | #define kernel extern "C" __declspec( dllexport ) |
---|
18 | static double trunc(double x) { return x>=0?floor(x):-floor(-x); } |
---|
19 | static double fmin(double x, double y) { return x>y ? y : x; } |
---|
20 | static double fmax(double x, double y) { return x<y ? y : x; } |
---|
21 | static double isnan(double x) { return _isnan(x); } |
---|
22 | #define NAN (std::numeric_limits<double>::quiet_NaN()) // non-signalling NaN |
---|
23 | static double cephes_expm1(double x) { |
---|
24 | // Adapted from the cephes math library. |
---|
25 | // Copyright 1984 - 1992 by Stephen L. Moshier |
---|
26 | if (x != x || x == 0.0) { |
---|
27 | return x; // NaN and +/- 0 |
---|
28 | } else if (x < -0.5 || x > 0.5) { |
---|
29 | return exp(x) - 1.0; |
---|
30 | } else { |
---|
31 | const double xsq = x*x; |
---|
32 | const double p = ((( |
---|
33 | +1.2617719307481059087798E-4)*xsq |
---|
34 | +3.0299440770744196129956E-2)*xsq |
---|
35 | +9.9999999999999999991025E-1); |
---|
36 | const double q = (((( |
---|
37 | +3.0019850513866445504159E-6)*xsq |
---|
38 | +2.5244834034968410419224E-3)*xsq |
---|
39 | +2.2726554820815502876593E-1)*xsq |
---|
40 | +2.0000000000000000000897E0); |
---|
41 | double r = x * p; |
---|
42 | r = r / (q - r); |
---|
43 | return r+r; |
---|
44 | } |
---|
45 | } |
---|
46 | #define expm1 cephes_expm1 |
---|
47 | typedef __int32 int32_t |
---|
48 | #else |
---|
49 | #define kernel extern "C" |
---|
50 | #include <cstdint> |
---|
51 | #endif |
---|
52 | static void SINCOS(double angle, double &svar, double &cvar) { svar=sin(angle); cvar=cos(angle); } |
---|
53 | # else |
---|
54 | #include <inttypes.h> // C99 guarantees that int32_t types is here |
---|
55 | #include <stdio.h> |
---|
56 | #include <tgmath.h> // C99 type-generic math, so sin(float) => sinf |
---|
57 | // MSVC doesn't support C99, so no need for dllexport on C99 branch |
---|
58 | #define kernel |
---|
59 | #define SINCOS(angle,svar,cvar) do {const double _t_=angle; svar=sin(_t_);cvar=cos(_t_);} while (0) |
---|
60 | # endif |
---|
61 | # define global |
---|
62 | # define local |
---|
63 | # define constant const |
---|
64 | // OpenCL powr(a,b) = C99 pow(a,b), b >= 0 |
---|
65 | // OpenCL pown(a,b) = C99 pow(a,b), b integer |
---|
66 | # define powr(a,b) pow(a,b) |
---|
67 | # define pown(a,b) pow(a,b) |
---|
68 | #else |
---|
69 | typedef int int32_t; |
---|
70 | # if defined(USE_SINCOS) |
---|
71 | # define SINCOS(angle,svar,cvar) svar=sincos(angle,&cvar) |
---|
72 | # else |
---|
73 | # define SINCOS(angle,svar,cvar) do {const double _t_=angle; svar=sin(_t_);cvar=cos(_t_);} while (0) |
---|
74 | # endif |
---|
75 | #endif |
---|
76 | |
---|
77 | // Standard mathematical constants: |
---|
78 | // M_E, M_LOG2E, M_LOG10E, M_LN2, M_LN10, M_PI, M_PI_2=pi/2, M_PI_4=pi/4, |
---|
79 | // M_1_PI=1/pi, M_2_PI=2/pi, M_2_SQRTPI=2/sqrt(pi), SQRT2, SQRT1_2=sqrt(1/2) |
---|
80 | // OpenCL defines M_constant_F for float constants, and nothing if double |
---|
81 | // is not enabled on the card, which is why these constants may be missing |
---|
82 | #ifndef M_PI |
---|
83 | # define M_PI 3.141592653589793 |
---|
84 | #endif |
---|
85 | #ifndef M_PI_2 |
---|
86 | # define M_PI_2 1.570796326794897 |
---|
87 | #endif |
---|
88 | #ifndef M_PI_4 |
---|
89 | # define M_PI_4 0.7853981633974483 |
---|
90 | #endif |
---|
91 | #ifndef M_E |
---|
92 | # define M_E 2.718281828459045091 |
---|
93 | #endif |
---|
94 | |
---|
95 | // Non-standard function library |
---|
96 | // pi/180, used for converting between degrees and radians |
---|
97 | // 4/3 pi for computing sphere volumes |
---|
98 | // square and cube for computing squares and cubes |
---|
99 | #ifndef M_PI_180 |
---|
100 | # define M_PI_180 0.017453292519943295 |
---|
101 | #endif |
---|
102 | #ifndef M_4PI_3 |
---|
103 | # define M_4PI_3 4.18879020478639 |
---|
104 | #endif |
---|
105 | static double square(double x) { return x*x; } |
---|
106 | static double cube(double x) { return x*x*x; } |
---|
107 | static double sinc(double x) { return x==0 ? 1.0 : sin(x)/x; } |
---|
108 | |
---|