File size: 3,181 Bytes
be11144
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
#pragma once

#ifdef __NVCC__ 
    #define DEVICE __device__ __host__ 
#else
    #define DEVICE
#endif

#ifndef __NVCC__
    #include <cmath>
    namespace {
        inline float fmodf(float a, float b) {
            return std::fmod(a, b);
        }
        inline double fmod(double a, double b) {
            return std::fmod(a, b);
        }
    }
    using std::isfinite;
#endif

#ifndef M_PI
#define M_PI 3.14159265358979323846
#endif

#include <cstdint>
#include <atomic>

// We use Real for most of the internal computation.
// However, for PyTorch interfaces, Optix Prime and Embree queries
// we use float
using Real = float;

template <typename T>
DEVICE
inline T square(const T &x) {
    return x * x;
}

template <typename T>
DEVICE
inline T cubic(const T &x) {
    return x * x * x;
}

template <typename T>
DEVICE
inline T clamp(const T &v, const T &lo, const T &hi) {
    if (v < lo) return lo;
    else if (v > hi) return hi;
    else return v;
}

DEVICE
inline int modulo(int a, int b) {
    auto r = a % b;
    return (r < 0) ? r+b : r;
}

DEVICE
inline float modulo(float a, float b) {
    float r = ::fmodf(a, b);
    return (r < 0.0f) ? r+b : r;
}

DEVICE
inline double modulo(double a, double b) {
    double r = ::fmod(a, b);
    return (r < 0.0) ? r+b : r;
}

template <typename T>
DEVICE
inline T max(const T &a, const T &b) {
    return a > b ? a : b;
}

template <typename T>
DEVICE
inline T min(const T &a, const T &b) {
    return a < b ? a : b;
}

/// Return ceil(x/y) for integers x and y
inline int idiv_ceil(int x, int y) {
    return (x + y-1) / y;
}

template <typename T>
DEVICE
inline void swap_(T &a, T &b) {
    T tmp = a;
    a = b;
    b = tmp;
}

inline double log2(double x) {
    return log(x) / log(Real(2));
}

template <typename T>
DEVICE
inline T safe_acos(const T &x) {
    if (x >= 1) return T(0);
    else if(x <= -1) return T(M_PI);
    return acos(x);
}

// For Morton code computation. This can be made faster.
DEVICE
inline uint32_t expand_bits(uint32_t x) {
    // Insert one zero after every bit given a 10-bit integer
    constexpr uint64_t mask = 0x1u;
    // We start from LSB (bit 31)
    auto result = (x & (mask << 0u));
    result |= ((x & (mask << 1u)) << 1u);
    result |= ((x & (mask << 2u)) << 2u);
    result |= ((x & (mask << 3u)) << 3u);
    result |= ((x & (mask << 4u)) << 4u);
    result |= ((x & (mask << 5u)) << 5u);
    result |= ((x & (mask << 6u)) << 6u);
    result |= ((x & (mask << 7u)) << 7u);
    result |= ((x & (mask << 8u)) << 8u);
    result |= ((x & (mask << 9u)) << 9u);
    return result;
}

// DEVICE
// inline int clz(uint64_t x) {
// #ifdef __CUDA_ARCH__
//     return __clzll(x);
// #else
//     // TODO: use _BitScanReverse in windows
//     return x == 0 ? 64 : __builtin_clzll(x);
// #endif
// }

// DEVICE
// inline int ffs(uint8_t x) {
// #ifdef __CUDA_ARCH__
//     return __ffs(x);
// #else
//     // TODO: use _BitScanReverse in windows
//     return __builtin_ffs(x);
// #endif
// }

// DEVICE
// inline int popc(uint8_t x) {
// #ifdef __CUDA_ARCH__
//     return __popc(x);
// #else
//     // TODO: use _popcnt in windows
//     return __builtin_popcount(x);
// #endif
// }