1 /// @ref simd
2 /// @file glm/simd/geometric.h
3
4 #pragma once
5
6 #include "common.h"
7
8 #if GLM_ARCH & GLM_ARCH_SSE2_BIT
9
10 GLM_FUNC_DECL glm_vec4 glm_vec4_dot(glm_vec4 v1, glm_vec4 v2);
11 GLM_FUNC_DECL glm_vec4 glm_vec1_dot(glm_vec4 v1, glm_vec4 v2);
12
glm_vec4_length(glm_vec4 x)13 GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_length(glm_vec4 x)
14 {
15 glm_vec4 const dot0 = glm_vec4_dot(x, x);
16 glm_vec4 const sqt0 = _mm_sqrt_ps(dot0);
17 return sqt0;
18 }
19
glm_vec4_distance(glm_vec4 p0,glm_vec4 p1)20 GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_distance(glm_vec4 p0, glm_vec4 p1)
21 {
22 glm_vec4 const sub0 = _mm_sub_ps(p0, p1);
23 glm_vec4 const len0 = glm_vec4_length(sub0);
24 return len0;
25 }
26
glm_vec4_dot(glm_vec4 v1,glm_vec4 v2)27 GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_dot(glm_vec4 v1, glm_vec4 v2)
28 {
29 # if GLM_ARCH & GLM_ARCH_AVX_BIT
30 return _mm_dp_ps(v1, v2, 0xff);
31 # elif GLM_ARCH & GLM_ARCH_SSE3_BIT
32 glm_vec4 const mul0 = _mm_mul_ps(v1, v2);
33 glm_vec4 const hadd0 = _mm_hadd_ps(mul0, mul0);
34 glm_vec4 const hadd1 = _mm_hadd_ps(hadd0, hadd0);
35 return hadd1;
36 # else
37 glm_vec4 const mul0 = _mm_mul_ps(v1, v2);
38 glm_vec4 const swp0 = _mm_shuffle_ps(mul0, mul0, _MM_SHUFFLE(2, 3, 0, 1));
39 glm_vec4 const add0 = _mm_add_ps(mul0, swp0);
40 glm_vec4 const swp1 = _mm_shuffle_ps(add0, add0, _MM_SHUFFLE(0, 1, 2, 3));
41 glm_vec4 const add1 = _mm_add_ps(add0, swp1);
42 return add1;
43 # endif
44 }
45
glm_vec1_dot(glm_vec4 v1,glm_vec4 v2)46 GLM_FUNC_QUALIFIER glm_vec4 glm_vec1_dot(glm_vec4 v1, glm_vec4 v2)
47 {
48 # if GLM_ARCH & GLM_ARCH_AVX_BIT
49 return _mm_dp_ps(v1, v2, 0xff);
50 # elif GLM_ARCH & GLM_ARCH_SSE3_BIT
51 glm_vec4 const mul0 = _mm_mul_ps(v1, v2);
52 glm_vec4 const had0 = _mm_hadd_ps(mul0, mul0);
53 glm_vec4 const had1 = _mm_hadd_ps(had0, had0);
54 return had1;
55 # else
56 glm_vec4 const mul0 = _mm_mul_ps(v1, v2);
57 glm_vec4 const mov0 = _mm_movehl_ps(mul0, mul0);
58 glm_vec4 const add0 = _mm_add_ps(mov0, mul0);
59 glm_vec4 const swp1 = _mm_shuffle_ps(add0, add0, 1);
60 glm_vec4 const add1 = _mm_add_ss(add0, swp1);
61 return add1;
62 # endif
63 }
64
glm_vec4_cross(glm_vec4 v1,glm_vec4 v2)65 GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_cross(glm_vec4 v1, glm_vec4 v2)
66 {
67 glm_vec4 const swp0 = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(3, 0, 2, 1));
68 glm_vec4 const swp1 = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(3, 1, 0, 2));
69 glm_vec4 const swp2 = _mm_shuffle_ps(v2, v2, _MM_SHUFFLE(3, 0, 2, 1));
70 glm_vec4 const swp3 = _mm_shuffle_ps(v2, v2, _MM_SHUFFLE(3, 1, 0, 2));
71 glm_vec4 const mul0 = _mm_mul_ps(swp0, swp3);
72 glm_vec4 const mul1 = _mm_mul_ps(swp1, swp2);
73 glm_vec4 const sub0 = _mm_sub_ps(mul0, mul1);
74 return sub0;
75 }
76
glm_vec4_normalize(glm_vec4 v)77 GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_normalize(glm_vec4 v)
78 {
79 glm_vec4 const dot0 = glm_vec4_dot(v, v);
80 glm_vec4 const isr0 = _mm_rsqrt_ps(dot0);
81 glm_vec4 const mul0 = _mm_mul_ps(v, isr0);
82 return mul0;
83 }
84
glm_vec4_faceforward(glm_vec4 N,glm_vec4 I,glm_vec4 Nref)85 GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_faceforward(glm_vec4 N, glm_vec4 I, glm_vec4 Nref)
86 {
87 glm_vec4 const dot0 = glm_vec4_dot(Nref, I);
88 glm_vec4 const sgn0 = glm_vec4_sign(dot0);
89 glm_vec4 const mul0 = _mm_mul_ps(sgn0, _mm_set1_ps(-1.0f));
90 glm_vec4 const mul1 = _mm_mul_ps(N, mul0);
91 return mul1;
92 }
93
glm_vec4_reflect(glm_vec4 I,glm_vec4 N)94 GLM_FUNC_QUALIFIER glm_vec4 glm_vec4_reflect(glm_vec4 I, glm_vec4 N)
95 {
96 glm_vec4 const dot0 = glm_vec4_dot(N, I);
97 glm_vec4 const mul0 = _mm_mul_ps(N, dot0);
98 glm_vec4 const mul1 = _mm_mul_ps(mul0, _mm_set1_ps(2.0f));
99 glm_vec4 const sub0 = _mm_sub_ps(I, mul1);
100 return sub0;
101 }
102
glm_vec4_refract(glm_vec4 I,glm_vec4 N,glm_vec4 eta)103 GLM_FUNC_QUALIFIER __m128 glm_vec4_refract(glm_vec4 I, glm_vec4 N, glm_vec4 eta)
104 {
105 glm_vec4 const dot0 = glm_vec4_dot(N, I);
106 glm_vec4 const mul0 = _mm_mul_ps(eta, eta);
107 glm_vec4 const mul1 = _mm_mul_ps(dot0, dot0);
108 glm_vec4 const sub0 = _mm_sub_ps(_mm_set1_ps(1.0f), mul0);
109 glm_vec4 const sub1 = _mm_sub_ps(_mm_set1_ps(1.0f), mul1);
110 glm_vec4 const mul2 = _mm_mul_ps(sub0, sub1);
111
112 if(_mm_movemask_ps(_mm_cmplt_ss(mul2, _mm_set1_ps(0.0f))) == 0)
113 return _mm_set1_ps(0.0f);
114
115 glm_vec4 const sqt0 = _mm_sqrt_ps(mul2);
116 glm_vec4 const mad0 = glm_vec4_fma(eta, dot0, sqt0);
117 glm_vec4 const mul4 = _mm_mul_ps(mad0, N);
118 glm_vec4 const mul5 = _mm_mul_ps(eta, I);
119 glm_vec4 const sub2 = _mm_sub_ps(mul5, mul4);
120
121 return sub2;
122 }
123
124 #endif//GLM_ARCH & GLM_ARCH_SSE2_BIT
125