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
|
/*
* Copyright (c) 2021, Stephan Unverwerth <s.unverwerth@serenityos.org>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#pragma once
#include <AK/SIMDExtras.h>
#include <LibGfx/Vector2.h>
#include <LibGfx/Vector3.h>
#include <LibGfx/Vector4.h>
namespace SoftGPU {
ALWAYS_INLINE static constexpr Vector2<AK::SIMD::f32x4> expand4(Vector2<float> const& v)
{
return Vector2<AK::SIMD::f32x4> {
AK::SIMD::expand4(v.x()),
AK::SIMD::expand4(v.y()),
};
}
ALWAYS_INLINE static constexpr Vector3<AK::SIMD::f32x4> expand4(Vector3<float> const& v)
{
return Vector3<AK::SIMD::f32x4> {
AK::SIMD::expand4(v.x()),
AK::SIMD::expand4(v.y()),
AK::SIMD::expand4(v.z()),
};
}
ALWAYS_INLINE static constexpr Vector4<AK::SIMD::f32x4> expand4(Vector4<float> const& v)
{
return Vector4<AK::SIMD::f32x4> {
AK::SIMD::expand4(v.x()),
AK::SIMD::expand4(v.y()),
AK::SIMD::expand4(v.z()),
AK::SIMD::expand4(v.w()),
};
}
ALWAYS_INLINE static constexpr Vector2<AK::SIMD::i32x4> expand4(Vector2<int> const& v)
{
return Vector2<AK::SIMD::i32x4> {
AK::SIMD::expand4(v.x()),
AK::SIMD::expand4(v.y()),
};
}
ALWAYS_INLINE static constexpr Vector3<AK::SIMD::i32x4> expand4(Vector3<int> const& v)
{
return Vector3<AK::SIMD::i32x4> {
AK::SIMD::expand4(v.x()),
AK::SIMD::expand4(v.y()),
AK::SIMD::expand4(v.z()),
};
}
ALWAYS_INLINE static constexpr Vector4<AK::SIMD::i32x4> expand4(Vector4<int> const& v)
{
return Vector4<AK::SIMD::i32x4> {
AK::SIMD::expand4(v.x()),
AK::SIMD::expand4(v.y()),
AK::SIMD::expand4(v.z()),
AK::SIMD::expand4(v.w()),
};
}
ALWAYS_INLINE static AK::SIMD::f32x4 ddx(AK::SIMD::f32x4 v)
{
return AK::SIMD::f32x4 {
v[1] - v[0],
v[1] - v[0],
v[3] - v[2],
v[3] - v[2],
};
}
ALWAYS_INLINE static AK::SIMD::f32x4 ddy(AK::SIMD::f32x4 v)
{
return AK::SIMD::f32x4 {
v[2] - v[0],
v[3] - v[1],
v[2] - v[0],
v[3] - v[1],
};
}
ALWAYS_INLINE static Vector2<AK::SIMD::f32x4> ddx(Vector2<AK::SIMD::f32x4> const& v)
{
return {
ddx(v.x()),
ddx(v.y()),
};
}
ALWAYS_INLINE static Vector2<AK::SIMD::f32x4> ddy(Vector2<AK::SIMD::f32x4> const& v)
{
return {
ddy(v.x()),
ddy(v.y()),
};
}
ALWAYS_INLINE static AK::SIMD::f32x4 length(Vector2<AK::SIMD::f32x4> const& v)
{
return AK::SIMD::sqrt(v.dot(v));
}
// Calculates a quadratic approximation of log2, exploiting the fact that IEEE754 floats are represented as mantissa * 2^exponent.
// See https://stackoverflow.com/questions/9411823/fast-log2float-x-implementation-c
ALWAYS_INLINE static AK::SIMD::f32x4 log2_approximate(AK::SIMD::f32x4 v)
{
union {
AK::SIMD::f32x4 float_val;
AK::SIMD::i32x4 int_val;
} u { v };
// Extract just the exponent minus 1, giving a lower integral bound for log2.
auto log = AK::SIMD::to_f32x4(((u.int_val >> 23) & 255) - 128);
// Replace the exponent with 0, giving a value between 1 and 2.
u.int_val &= ~(255 << 23);
u.int_val |= 127 << 23;
// Approximate log2 by adding a quadratic function of u to the integral part.
log += (-0.34484843f * u.float_val + 2.02466578f) * u.float_val - 0.67487759f;
return log;
}
ALWAYS_INLINE static Vector2<AK::SIMD::f32x4> to_vec2_f32x4(Vector2<AK::SIMD::i32x4> const& v)
{
return {
AK::SIMD::to_f32x4(v.x()),
AK::SIMD::to_f32x4(v.y()),
};
}
}
|
