| 1 | /// @ref core |
| 2 | /// @file glm/detail/func_common_simd.inl |
| 3 | |
| 4 | #if GLM_ARCH & GLM_ARCH_SSE2_BIT |
| 5 | |
| 6 | #include "../simd/common.h" |
| 7 | |
| 8 | #include <immintrin.h> |
| 9 | |
| 10 | namespace glm{ |
| 11 | namespace detail |
| 12 | { |
| 13 | template<qualifier Q> |
| 14 | struct compute_abs_vector<4, float, Q, true> |
| 15 | { |
| 16 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v) |
| 17 | { |
| 18 | vec<4, float, Q> result; |
| 19 | result.data = glm_vec4_abs(v.data); |
| 20 | return result; |
| 21 | } |
| 22 | }; |
| 23 | |
| 24 | template<qualifier Q> |
| 25 | struct compute_abs_vector<4, int, Q, true> |
| 26 | { |
| 27 | GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& v) |
| 28 | { |
| 29 | vec<4, int, Q> result; |
| 30 | result.data = glm_ivec4_abs(v.data); |
| 31 | return result; |
| 32 | } |
| 33 | }; |
| 34 | |
| 35 | template<qualifier Q> |
| 36 | struct compute_floor<4, float, Q, true> |
| 37 | { |
| 38 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v) |
| 39 | { |
| 40 | vec<4, float, Q> result; |
| 41 | result.data = glm_vec4_floor(v.data); |
| 42 | return result; |
| 43 | } |
| 44 | }; |
| 45 | |
| 46 | template<qualifier Q> |
| 47 | struct compute_ceil<4, float, Q, true> |
| 48 | { |
| 49 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v) |
| 50 | { |
| 51 | vec<4, float, Q> result; |
| 52 | result.data = glm_vec4_ceil(v.data); |
| 53 | return result; |
| 54 | } |
| 55 | }; |
| 56 | |
| 57 | template<qualifier Q> |
| 58 | struct compute_fract<4, float, Q, true> |
| 59 | { |
| 60 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v) |
| 61 | { |
| 62 | vec<4, float, Q> result; |
| 63 | result.data = glm_vec4_fract(v.data); |
| 64 | return result; |
| 65 | } |
| 66 | }; |
| 67 | |
| 68 | template<qualifier Q> |
| 69 | struct compute_round<4, float, Q, true> |
| 70 | { |
| 71 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v) |
| 72 | { |
| 73 | vec<4, float, Q> result; |
| 74 | result.data = glm_vec4_round(v.data); |
| 75 | return result; |
| 76 | } |
| 77 | }; |
| 78 | |
| 79 | template<qualifier Q> |
| 80 | struct compute_mod<4, float, Q, true> |
| 81 | { |
| 82 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& x, vec<4, float, Q> const& y) |
| 83 | { |
| 84 | vec<4, float, Q> result; |
| 85 | result.data = glm_vec4_mod(x.data, y.data); |
| 86 | return result; |
| 87 | } |
| 88 | }; |
| 89 | |
| 90 | template<qualifier Q> |
| 91 | struct compute_min_vector<4, float, Q, true> |
| 92 | { |
| 93 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2) |
| 94 | { |
| 95 | vec<4, float, Q> result; |
| 96 | result.data = _mm_min_ps(v1.data, v2.data); |
| 97 | return result; |
| 98 | } |
| 99 | }; |
| 100 | |
| 101 | template<qualifier Q> |
| 102 | struct compute_min_vector<4, int, Q, true> |
| 103 | { |
| 104 | GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2) |
| 105 | { |
| 106 | vec<4, int, Q> result; |
| 107 | result.data = _mm_min_epi32(v1.data, v2.data); |
| 108 | return result; |
| 109 | } |
| 110 | }; |
| 111 | |
| 112 | template<qualifier Q> |
| 113 | struct compute_min_vector<4, uint, Q, true> |
| 114 | { |
| 115 | GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v1, vec<4, uint, Q> const& v2) |
| 116 | { |
| 117 | vec<4, uint, Q> result; |
| 118 | result.data = _mm_min_epu32(v1.data, v2.data); |
| 119 | return result; |
| 120 | } |
| 121 | }; |
| 122 | |
| 123 | template<qualifier Q> |
| 124 | struct compute_max_vector<4, float, Q, true> |
| 125 | { |
| 126 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& v1, vec<4, float, Q> const& v2) |
| 127 | { |
| 128 | vec<4, float, Q> result; |
| 129 | result.data = _mm_max_ps(v1.data, v2.data); |
| 130 | return result; |
| 131 | } |
| 132 | }; |
| 133 | |
| 134 | template<qualifier Q> |
| 135 | struct compute_max_vector<4, int, Q, true> |
| 136 | { |
| 137 | GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& v1, vec<4, int, Q> const& v2) |
| 138 | { |
| 139 | vec<4, int, Q> result; |
| 140 | result.data = _mm_max_epi32(v1.data, v2.data); |
| 141 | return result; |
| 142 | } |
| 143 | }; |
| 144 | |
| 145 | template<qualifier Q> |
| 146 | struct compute_max_vector<4, uint, Q, true> |
| 147 | { |
| 148 | GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& v1, vec<4, uint, Q> const& v2) |
| 149 | { |
| 150 | vec<4, uint, Q> result; |
| 151 | result.data = _mm_max_epu32(v1.data, v2.data); |
| 152 | return result; |
| 153 | } |
| 154 | }; |
| 155 | |
| 156 | template<qualifier Q> |
| 157 | struct compute_clamp_vector<4, float, Q, true> |
| 158 | { |
| 159 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& x, vec<4, float, Q> const& minVal, vec<4, float, Q> const& maxVal) |
| 160 | { |
| 161 | vec<4, float, Q> result; |
| 162 | result.data = _mm_min_ps(_mm_max_ps(x.data, minVal.data), maxVal.data); |
| 163 | return result; |
| 164 | } |
| 165 | }; |
| 166 | |
| 167 | template<qualifier Q> |
| 168 | struct compute_clamp_vector<4, int, Q, true> |
| 169 | { |
| 170 | GLM_FUNC_QUALIFIER static vec<4, int, Q> call(vec<4, int, Q> const& x, vec<4, int, Q> const& minVal, vec<4, int, Q> const& maxVal) |
| 171 | { |
| 172 | vec<4, int, Q> result; |
| 173 | result.data = _mm_min_epi32(_mm_max_epi32(x.data, minVal.data), maxVal.data); |
| 174 | return result; |
| 175 | } |
| 176 | }; |
| 177 | |
| 178 | template<qualifier Q> |
| 179 | struct compute_clamp_vector<4, uint, Q, true> |
| 180 | { |
| 181 | GLM_FUNC_QUALIFIER static vec<4, uint, Q> call(vec<4, uint, Q> const& x, vec<4, uint, Q> const& minVal, vec<4, uint, Q> const& maxVal) |
| 182 | { |
| 183 | vec<4, uint, Q> result; |
| 184 | result.data = _mm_min_epu32(_mm_max_epu32(x.data, minVal.data), maxVal.data); |
| 185 | return result; |
| 186 | } |
| 187 | }; |
| 188 | |
| 189 | template<qualifier Q> |
| 190 | struct compute_mix_vector<4, float, bool, Q, true> |
| 191 | { |
| 192 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& x, vec<4, float, Q> const& y, vec<4, bool, Q> const& a) |
| 193 | { |
| 194 | __m128i const Load = _mm_set_epi32(-static_cast<int>(a.w), -static_cast<int>(a.z), -static_cast<int>(a.y), -static_cast<int>(a.x)); |
| 195 | __m128 const Mask = _mm_castsi128_ps(Load); |
| 196 | |
| 197 | vec<4, float, Q> Result; |
| 198 | # if 0 && GLM_ARCH & GLM_ARCH_AVX |
| 199 | Result.data = _mm_blendv_ps(x.data, y.data, Mask); |
| 200 | # else |
| 201 | Result.data = _mm_or_ps(_mm_and_ps(Mask, y.data), _mm_andnot_ps(Mask, x.data)); |
| 202 | # endif |
| 203 | return Result; |
| 204 | } |
| 205 | }; |
| 206 | /* FIXME |
| 207 | template<qualifier Q> |
| 208 | struct compute_step_vector<float, Q, tvec4> |
| 209 | { |
| 210 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& edge, vec<4, float, Q> const& x) |
| 211 | { |
| 212 | vec<4, float, Q> Result; |
| 213 | result.data = glm_vec4_step(edge.data, x.data); |
| 214 | return result; |
| 215 | } |
| 216 | }; |
| 217 | */ |
| 218 | template<qualifier Q> |
| 219 | struct compute_smoothstep_vector<4, float, Q, true> |
| 220 | { |
| 221 | GLM_FUNC_QUALIFIER static vec<4, float, Q> call(vec<4, float, Q> const& edge0, vec<4, float, Q> const& edge1, vec<4, float, Q> const& x) |
| 222 | { |
| 223 | vec<4, float, Q> Result; |
| 224 | Result.data = glm_vec4_smoothstep(edge0.data, edge1.data, x.data); |
| 225 | return Result; |
| 226 | } |
| 227 | }; |
| 228 | }//namespace detail |
| 229 | }//namespace glm |
| 230 | |
| 231 | #endif//GLM_ARCH & GLM_ARCH_SSE2_BIT |