llml 0.2.3

Implementation of basic math data types with high level frontend and low level backend
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

x86_use!();

use cfg_if::cfg_if;
use std::{ops::{Add, Sub, Mul, Div, Neg}};
use crate::{x86::{_mm_low_ps, _mm_high_ps}, others::Zero};
use super::{vec4::EucVecf4, vec2::EucVecf2, Matd2};

#[derive(Clone, Copy, PartialEq, Eq)]
#[repr(transparent)]
pub struct Matf2 (pub(crate) EucVecf4);

impl Matf2 {
    #[inline]
    pub fn new (a: [f32;4]) -> Self {
        Self(a.into())
    }

    #[inline]
    pub fn of_rot (a: f32) -> Self {
        let (sin, cos) = a.sin_cos();
        Self::new([cos, -sin, sin, cos])
    }

    #[inline(always)]
    pub fn x (&self) -> EucVecf2 {
        unsafe { EucVecf2(_mm_high_ps(self.0.0)) }
    }

    #[inline(always)]
    pub fn xx (&self) -> f32 {
        self.0.x()
    }

    #[inline(always)]
    pub fn xy (&self) -> f32 {
        self.0.y()
    }

    #[inline(always)]
    pub fn y (&self) -> EucVecf2 {
        unsafe { EucVecf2(_mm_low_ps(self.0.0)) }
    }

    #[inline(always)]
    pub fn yx (&self) -> f32 {
        self.0.z()
    }

    #[inline(always)]
    pub fn yy (&self) -> f32 {
        self.0.w()
    }

    #[inline(always)]
    pub fn scal_mul (self, rhs: Self) -> Self {
        Self(self.0 * rhs.0)
    }

    #[inline(always)]
    pub fn scal_div (self, rhs: Self) -> Self {
        Self(self.0 / rhs.0)
    }

    #[inline(always)]
    pub fn tr (self) -> f32 {
        self.0.x() + self.0.w()
    }

    #[inline(always)]
    pub fn det (self) -> f32 {
        unsafe {
            let v2 = _mm_shuffle_ps(self.0.0, _mm_setzero_ps(), _MM_SHUFFLE(0, 0, 2, 3));
            let m1 = EucVecf2(_mm_mul_ps(self.0.0, v2));

            m1.x() - m1.y()
        }
    }

    #[inline(always)]
    pub fn inv (self) -> Option<Self> {
        let det = self.det();
        if det.is_zero() {
            return None
        }

        unsafe { Some(self._inv(det)) }
    }

    #[inline(always)]
    pub unsafe fn inv_unsafe (self) -> Self {
        self._inv(self.det())
    }

    #[inline(always)]
    unsafe fn _inv (self, det: f32) -> Self {
        let neg = EucVecf4(_mm_sub_ps(_mm_setzero_ps(), _mm_shuffle_ps(self.0.0, self.0.0, _MM_SHUFFLE(0, 0, 2, 1))));
        Self(EucVecf4::new([self.0.w(), neg.x(), neg.y(), self.0.x()]) / det)
    }
}

trait_map!(
    Matf2, f32,
    Add, add,
    Sub, sub
);

trait_map_scal!(
    Matf2, f32,
    Mul, mul,
    Div, div
);

impl Neg for Matf2 {
    type Output = Self;

    #[inline(always)]
    fn neg(self) -> Self::Output {
        Self(-self.0)
    }
}

impl Mul<EucVecf2> for Matf2 {
    type Output = EucVecf2;

    #[inline(always)]
    fn mul (self, rhs: EucVecf2) -> Self::Output {
        unsafe {
            let v1 = self.0.0;
            let v2 = _mm_shuffle_ps(rhs.0, rhs.0, _MM_SHUFFLE(1, 0, 1, 0));
            let m1 = _mm_mul_ps(v1, v2);
            
            let v1 : __m128;
            cfg_if! {
                if #[cfg(target_feature = "sse3")] {
                    v1 = _mm_moveldup_ps(m1)
                } else {
                    v1 = _mm_shuffle_ps(m1, m1, _MM_SHUFFLE(2, 2, 0, 0));
                }
            }

            EucVecf2(_mm_shuffle_ps(_mm_add_ps(m1, v1), _mm_setzero_ps(), _MM_SHUFFLE(0, 0, 3, 1)))
        }
    }
}

impl Mul for Matf2 {
    type Output = Self;

    #[inline(always)]
    fn mul(self, rhs: Self) -> Self::Output {
        unsafe {
            let v1 : __m128;
            let v3 : __m128;

            cfg_if! {
                if #[cfg(target_feature = "sse3")] {
                    v1 = _mm_moveldup_ps(self.0.0);
                    v3 = _mm_movehdup_ps(self.0.0);
                } else {
                    v1 = _mm_shuffle_ps(self.0.0, self.0.0, _MM_SHUFFLE(2, 2, 0, 0));
                    v3 = _mm_shuffle_ps(self.0.0, self.0.0, _MM_SHUFFLE(3, 3, 1, 1));
                }
            }
            
            let v2 = _mm_shuffle_ps(rhs.0.0, rhs.0.0, _MM_SHUFFLE(1, 0, 1, 0));
            let v4 = _mm_shuffle_ps(rhs.0.0, rhs.0.0, _MM_SHUFFLE(3, 2, 3, 2));

            let m1 = _mm_mul_ps(v1, v2);
            let m2 = _mm_mul_ps(v3, v4);
            Self(EucVecf4(_mm_add_ps(m1, m2)))
        }
    }
}

impl Into<[f32;4]> for Matf2 {
    #[inline(always)]
    fn into(self) -> [f32;4] {
        self.0.into()
    }
}

impl Into<Matd2> for Matf2 {
    #[inline(always)]
    fn into(self) -> Matd2 {
        Matd2(self.0.into())
    }
}