1 /////////////////////////////////////////////////////////////////////////////////// 2 /// OpenGL Mathematics (glm.g-truc.net) 3 /// 4 /// Copyright (c) 2005 - 2014 G-Truc Creation (www.g-truc.net) 5 /// Permission is hereby granted, free of charge, to any person obtaining a copy 6 /// of this software and associated documentation files (the "Software"), to deal 7 /// in the Software without restriction, including without limitation the rights 8 /// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 9 /// copies of the Software, and to permit persons to whom the Software is 10 /// furnished to do so, subject to the following conditions: 11 /// 12 /// The above copyright notice and this permission notice shall be included in 13 /// all copies or substantial portions of the Software. 14 /// 15 /// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 /// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 /// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 18 /// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 /// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 20 /// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 21 /// THE SOFTWARE. 22 /// 23 /// @ref core 24 /// @file glm/core/intrinsic_common.inl 25 /// @date 2009-06-05 / 2011-06-15 26 /// @author Christophe Riccio 27 /////////////////////////////////////////////////////////////////////////////////// 28 29 namespace glm{ 30 namespace detail{ 31 32 static const __m128 GLM_VAR_USED _m128_rad_ps = _mm_set_ps1(3.141592653589793238462643383279f / 180.f); 33 static const __m128 GLM_VAR_USED _m128_deg_ps = _mm_set_ps1(180.f / 3.141592653589793238462643383279f); 34 35 template <typename matType> 36 GLM_FUNC_QUALIFIER matType sse_comp_mul_ps 37 ( 38 __m128 const in1[4], 39 __m128 const in2[4], 40 __m128 out[4] 41 ) 42 { 43 out[0] = _mm_mul_ps(in1[0], in2[0]); 44 out[1] = _mm_mul_ps(in1[1], in2[1]); 45 out[2] = _mm_mul_ps(in1[2], in2[2]); 46 out[3] = _mm_mul_ps(in1[3], in2[3]); 47 } 48 49 GLM_FUNC_QUALIFIER void sse_add_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4]) 50 { 51 { 52 out[0] = _mm_add_ps(in1[0], in2[0]); 53 out[1] = _mm_add_ps(in1[1], in2[1]); 54 out[2] = _mm_add_ps(in1[2], in2[2]); 55 out[3] = _mm_add_ps(in1[3], in2[3]); 56 } 57 } 58 59 GLM_FUNC_QUALIFIER void sse_sub_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4]) 60 { 61 { 62 out[0] = _mm_sub_ps(in1[0], in2[0]); 63 out[1] = _mm_sub_ps(in1[1], in2[1]); 64 out[2] = _mm_sub_ps(in1[2], in2[2]); 65 out[3] = _mm_sub_ps(in1[3], in2[3]); 66 } 67 } 68 69 GLM_FUNC_QUALIFIER __m128 sse_mul_ps(__m128 const m[4], __m128 v) 70 { 71 __m128 v0 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0)); 72 __m128 v1 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1)); 73 __m128 v2 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2)); 74 __m128 v3 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3)); 75 76 __m128 m0 = _mm_mul_ps(m[0], v0); 77 __m128 m1 = _mm_mul_ps(m[1], v1); 78 __m128 m2 = _mm_mul_ps(m[2], v2); 79 __m128 m3 = _mm_mul_ps(m[3], v3); 80 81 __m128 a0 = _mm_add_ps(m0, m1); 82 __m128 a1 = _mm_add_ps(m2, m3); 83 __m128 a2 = _mm_add_ps(a0, a1); 84 85 return a2; 86 } 87 88 GLM_FUNC_QUALIFIER __m128 sse_mul_ps(__m128 v, __m128 const m[4]) 89 { 90 __m128 i0 = m[0]; 91 __m128 i1 = m[1]; 92 __m128 i2 = m[2]; 93 __m128 i3 = m[3]; 94 95 __m128 m0 = _mm_mul_ps(v, i0); 96 __m128 m1 = _mm_mul_ps(v, i1); 97 __m128 m2 = _mm_mul_ps(v, i2); 98 __m128 m3 = _mm_mul_ps(v, i3); 99 100 __m128 u0 = _mm_unpacklo_ps(m0, m1); 101 __m128 u1 = _mm_unpackhi_ps(m0, m1); 102 __m128 a0 = _mm_add_ps(u0, u1); 103 104 __m128 u2 = _mm_unpacklo_ps(m2, m3); 105 __m128 u3 = _mm_unpackhi_ps(m2, m3); 106 __m128 a1 = _mm_add_ps(u2, u3); 107 108 __m128 f0 = _mm_movelh_ps(a0, a1); 109 __m128 f1 = _mm_movehl_ps(a1, a0); 110 __m128 f2 = _mm_add_ps(f0, f1); 111 112 return f2; 113 } 114 115 GLM_FUNC_QUALIFIER void sse_mul_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4]) 116 { 117 { 118 __m128 e0 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(0, 0, 0, 0)); 119 __m128 e1 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(1, 1, 1, 1)); 120 __m128 e2 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(2, 2, 2, 2)); 121 __m128 e3 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(3, 3, 3, 3)); 122 123 __m128 m0 = _mm_mul_ps(in1[0], e0); 124 __m128 m1 = _mm_mul_ps(in1[1], e1); 125 __m128 m2 = _mm_mul_ps(in1[2], e2); 126 __m128 m3 = _mm_mul_ps(in1[3], e3); 127 128 __m128 a0 = _mm_add_ps(m0, m1); 129 __m128 a1 = _mm_add_ps(m2, m3); 130 __m128 a2 = _mm_add_ps(a0, a1); 131 132 out[0] = a2; 133 } 134 135 { 136 __m128 e0 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(0, 0, 0, 0)); 137 __m128 e1 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(1, 1, 1, 1)); 138 __m128 e2 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(2, 2, 2, 2)); 139 __m128 e3 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(3, 3, 3, 3)); 140 141 __m128 m0 = _mm_mul_ps(in1[0], e0); 142 __m128 m1 = _mm_mul_ps(in1[1], e1); 143 __m128 m2 = _mm_mul_ps(in1[2], e2); 144 __m128 m3 = _mm_mul_ps(in1[3], e3); 145 146 __m128 a0 = _mm_add_ps(m0, m1); 147 __m128 a1 = _mm_add_ps(m2, m3); 148 __m128 a2 = _mm_add_ps(a0, a1); 149 150 out[1] = a2; 151 } 152 153 { 154 __m128 e0 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(0, 0, 0, 0)); 155 __m128 e1 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(1, 1, 1, 1)); 156 __m128 e2 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(2, 2, 2, 2)); 157 __m128 e3 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(3, 3, 3, 3)); 158 159 __m128 m0 = _mm_mul_ps(in1[0], e0); 160 __m128 m1 = _mm_mul_ps(in1[1], e1); 161 __m128 m2 = _mm_mul_ps(in1[2], e2); 162 __m128 m3 = _mm_mul_ps(in1[3], e3); 163 164 __m128 a0 = _mm_add_ps(m0, m1); 165 __m128 a1 = _mm_add_ps(m2, m3); 166 __m128 a2 = _mm_add_ps(a0, a1); 167 168 out[2] = a2; 169 } 170 171 { 172 //(__m128&)_mm_shuffle_epi32(__m128i&)in2[0], _MM_SHUFFLE(3, 3, 3, 3)) 173 __m128 e0 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(0, 0, 0, 0)); 174 __m128 e1 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(1, 1, 1, 1)); 175 __m128 e2 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(2, 2, 2, 2)); 176 __m128 e3 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(3, 3, 3, 3)); 177 178 __m128 m0 = _mm_mul_ps(in1[0], e0); 179 __m128 m1 = _mm_mul_ps(in1[1], e1); 180 __m128 m2 = _mm_mul_ps(in1[2], e2); 181 __m128 m3 = _mm_mul_ps(in1[3], e3); 182 183 __m128 a0 = _mm_add_ps(m0, m1); 184 __m128 a1 = _mm_add_ps(m2, m3); 185 __m128 a2 = _mm_add_ps(a0, a1); 186 187 out[3] = a2; 188 } 189 } 190 191 GLM_FUNC_QUALIFIER void sse_transpose_ps(__m128 const in[4], __m128 out[4]) 192 { 193 __m128 tmp0 = _mm_shuffle_ps(in[0], in[1], 0x44); 194 __m128 tmp2 = _mm_shuffle_ps(in[0], in[1], 0xEE); 195 __m128 tmp1 = _mm_shuffle_ps(in[2], in[3], 0x44); 196 __m128 tmp3 = _mm_shuffle_ps(in[2], in[3], 0xEE); 197 198 out[0] = _mm_shuffle_ps(tmp0, tmp1, 0x88); 199 out[1] = _mm_shuffle_ps(tmp0, tmp1, 0xDD); 200 out[2] = _mm_shuffle_ps(tmp2, tmp3, 0x88); 201 out[3] = _mm_shuffle_ps(tmp2, tmp3, 0xDD); 202 } 203 204 GLM_FUNC_QUALIFIER __m128 sse_slow_det_ps(__m128 const in[4]) 205 { 206 __m128 Fac0; 207 { 208 // valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3]; 209 // valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3]; 210 // valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3]; 211 // valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3]; 212 213 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 214 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 215 216 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 217 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 218 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 219 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 220 221 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 222 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 223 Fac0 = _mm_sub_ps(Mul00, Mul01); 224 } 225 226 __m128 Fac1; 227 { 228 // valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3]; 229 // valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3]; 230 // valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3]; 231 // valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3]; 232 233 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 234 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 235 236 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 237 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 238 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 239 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 240 241 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 242 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 243 Fac1 = _mm_sub_ps(Mul00, Mul01); 244 } 245 246 247 __m128 Fac2; 248 { 249 // valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2]; 250 // valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2]; 251 // valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2]; 252 // valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2]; 253 254 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 255 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 256 257 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 258 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 259 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 260 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 261 262 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 263 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 264 Fac2 = _mm_sub_ps(Mul00, Mul01); 265 } 266 267 __m128 Fac3; 268 { 269 // valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3]; 270 // valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3]; 271 // valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3]; 272 // valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3]; 273 274 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 275 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 276 277 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 278 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 279 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 280 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 281 282 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 283 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 284 Fac3 = _mm_sub_ps(Mul00, Mul01); 285 } 286 287 __m128 Fac4; 288 { 289 // valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2]; 290 // valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2]; 291 // valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2]; 292 // valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2]; 293 294 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 295 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 296 297 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 298 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 299 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 300 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 301 302 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 303 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 304 Fac4 = _mm_sub_ps(Mul00, Mul01); 305 } 306 307 __m128 Fac5; 308 { 309 // valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1]; 310 // valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1]; 311 // valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1]; 312 // valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1]; 313 314 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 315 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 316 317 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 318 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 319 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 320 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 321 322 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 323 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 324 Fac5 = _mm_sub_ps(Mul00, Mul01); 325 } 326 327 __m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f); 328 __m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f); 329 330 // m[1][0] 331 // m[0][0] 332 // m[0][0] 333 // m[0][0] 334 __m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0)); 335 __m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0)); 336 337 // m[1][1] 338 // m[0][1] 339 // m[0][1] 340 // m[0][1] 341 __m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1)); 342 __m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0)); 343 344 // m[1][2] 345 // m[0][2] 346 // m[0][2] 347 // m[0][2] 348 __m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2)); 349 __m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0)); 350 351 // m[1][3] 352 // m[0][3] 353 // m[0][3] 354 // m[0][3] 355 __m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3)); 356 __m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0)); 357 358 // col0 359 // + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]), 360 // - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]), 361 // + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]), 362 // - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]), 363 __m128 Mul00 = _mm_mul_ps(Vec1, Fac0); 364 __m128 Mul01 = _mm_mul_ps(Vec2, Fac1); 365 __m128 Mul02 = _mm_mul_ps(Vec3, Fac2); 366 __m128 Sub00 = _mm_sub_ps(Mul00, Mul01); 367 __m128 Add00 = _mm_add_ps(Sub00, Mul02); 368 __m128 Inv0 = _mm_mul_ps(SignB, Add00); 369 370 // col1 371 // - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]), 372 // + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]), 373 // - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]), 374 // + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]), 375 __m128 Mul03 = _mm_mul_ps(Vec0, Fac0); 376 __m128 Mul04 = _mm_mul_ps(Vec2, Fac3); 377 __m128 Mul05 = _mm_mul_ps(Vec3, Fac4); 378 __m128 Sub01 = _mm_sub_ps(Mul03, Mul04); 379 __m128 Add01 = _mm_add_ps(Sub01, Mul05); 380 __m128 Inv1 = _mm_mul_ps(SignA, Add01); 381 382 // col2 383 // + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]), 384 // - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]), 385 // + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]), 386 // - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]), 387 __m128 Mul06 = _mm_mul_ps(Vec0, Fac1); 388 __m128 Mul07 = _mm_mul_ps(Vec1, Fac3); 389 __m128 Mul08 = _mm_mul_ps(Vec3, Fac5); 390 __m128 Sub02 = _mm_sub_ps(Mul06, Mul07); 391 __m128 Add02 = _mm_add_ps(Sub02, Mul08); 392 __m128 Inv2 = _mm_mul_ps(SignB, Add02); 393 394 // col3 395 // - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]), 396 // + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]), 397 // - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]), 398 // + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3])); 399 __m128 Mul09 = _mm_mul_ps(Vec0, Fac2); 400 __m128 Mul10 = _mm_mul_ps(Vec1, Fac4); 401 __m128 Mul11 = _mm_mul_ps(Vec2, Fac5); 402 __m128 Sub03 = _mm_sub_ps(Mul09, Mul10); 403 __m128 Add03 = _mm_add_ps(Sub03, Mul11); 404 __m128 Inv3 = _mm_mul_ps(SignA, Add03); 405 406 __m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0)); 407 __m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0)); 408 __m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0)); 409 410 // valType Determinant = m[0][0] * Inverse[0][0] 411 // + m[0][1] * Inverse[1][0] 412 // + m[0][2] * Inverse[2][0] 413 // + m[0][3] * Inverse[3][0]; 414 __m128 Det0 = sse_dot_ps(in[0], Row2); 415 return Det0; 416 } 417 418 GLM_FUNC_QUALIFIER __m128 sse_detd_ps 419 ( 420 __m128 const m[4] 421 ) 422 { 423 // _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128( 424 425 //T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3]; 426 //T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3]; 427 //T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2]; 428 //T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3]; 429 //T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2]; 430 //T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1]; 431 432 // First 2 columns 433 __m128 Swp2A = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(0, 1, 1, 2))); 434 __m128 Swp3A = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(3, 2, 3, 3))); 435 __m128 MulA = _mm_mul_ps(Swp2A, Swp3A); 436 437 // Second 2 columns 438 __m128 Swp2B = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(3, 2, 3, 3))); 439 __m128 Swp3B = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(0, 1, 1, 2))); 440 __m128 MulB = _mm_mul_ps(Swp2B, Swp3B); 441 442 // Columns subtraction 443 __m128 SubE = _mm_sub_ps(MulA, MulB); 444 445 // Last 2 rows 446 __m128 Swp2C = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(0, 0, 1, 2))); 447 __m128 Swp3C = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(1, 2, 0, 0))); 448 __m128 MulC = _mm_mul_ps(Swp2C, Swp3C); 449 __m128 SubF = _mm_sub_ps(_mm_movehl_ps(MulC, MulC), MulC); 450 451 //detail::tvec4<T, P> DetCof( 452 // + (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02), 453 // - (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04), 454 // + (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05), 455 // - (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05)); 456 457 __m128 SubFacA = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubE), _MM_SHUFFLE(2, 1, 0, 0))); 458 __m128 SwpFacA = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(0, 0, 0, 1))); 459 __m128 MulFacA = _mm_mul_ps(SwpFacA, SubFacA); 460 461 __m128 SubTmpB = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(0, 0, 3, 1)); 462 __m128 SubFacB = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubTmpB), _MM_SHUFFLE(3, 1, 1, 0)));//SubF[0], SubE[3], SubE[3], SubE[1]; 463 __m128 SwpFacB = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(1, 1, 2, 2))); 464 __m128 MulFacB = _mm_mul_ps(SwpFacB, SubFacB); 465 466 __m128 SubRes = _mm_sub_ps(MulFacA, MulFacB); 467 468 __m128 SubTmpC = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(1, 0, 2, 2)); 469 __m128 SubFacC = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubTmpC), _MM_SHUFFLE(3, 3, 2, 0))); 470 __m128 SwpFacC = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(2, 3, 3, 3))); 471 __m128 MulFacC = _mm_mul_ps(SwpFacC, SubFacC); 472 473 __m128 AddRes = _mm_add_ps(SubRes, MulFacC); 474 __m128 DetCof = _mm_mul_ps(AddRes, _mm_setr_ps( 1.0f,-1.0f, 1.0f,-1.0f)); 475 476 //return m[0][0] * DetCof[0] 477 // + m[0][1] * DetCof[1] 478 // + m[0][2] * DetCof[2] 479 // + m[0][3] * DetCof[3]; 480 481 return sse_dot_ps(m[0], DetCof); 482 } 483 484 GLM_FUNC_QUALIFIER __m128 sse_det_ps 485 ( 486 __m128 const m[4] 487 ) 488 { 489 // _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(add) 490 491 //T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3]; 492 //T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3]; 493 //T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2]; 494 //T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3]; 495 //T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2]; 496 //T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1]; 497 498 // First 2 columns 499 __m128 Swp2A = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(0, 1, 1, 2)); 500 __m128 Swp3A = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(3, 2, 3, 3)); 501 __m128 MulA = _mm_mul_ps(Swp2A, Swp3A); 502 503 // Second 2 columns 504 __m128 Swp2B = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(3, 2, 3, 3)); 505 __m128 Swp3B = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(0, 1, 1, 2)); 506 __m128 MulB = _mm_mul_ps(Swp2B, Swp3B); 507 508 // Columns subtraction 509 __m128 SubE = _mm_sub_ps(MulA, MulB); 510 511 // Last 2 rows 512 __m128 Swp2C = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(0, 0, 1, 2)); 513 __m128 Swp3C = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(1, 2, 0, 0)); 514 __m128 MulC = _mm_mul_ps(Swp2C, Swp3C); 515 __m128 SubF = _mm_sub_ps(_mm_movehl_ps(MulC, MulC), MulC); 516 517 //detail::tvec4<T, P> DetCof( 518 // + (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02), 519 // - (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04), 520 // + (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05), 521 // - (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05)); 522 523 __m128 SubFacA = _mm_shuffle_ps(SubE, SubE, _MM_SHUFFLE(2, 1, 0, 0)); 524 __m128 SwpFacA = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(0, 0, 0, 1)); 525 __m128 MulFacA = _mm_mul_ps(SwpFacA, SubFacA); 526 527 __m128 SubTmpB = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(0, 0, 3, 1)); 528 __m128 SubFacB = _mm_shuffle_ps(SubTmpB, SubTmpB, _MM_SHUFFLE(3, 1, 1, 0));//SubF[0], SubE[3], SubE[3], SubE[1]; 529 __m128 SwpFacB = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(1, 1, 2, 2)); 530 __m128 MulFacB = _mm_mul_ps(SwpFacB, SubFacB); 531 532 __m128 SubRes = _mm_sub_ps(MulFacA, MulFacB); 533 534 __m128 SubTmpC = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(1, 0, 2, 2)); 535 __m128 SubFacC = _mm_shuffle_ps(SubTmpC, SubTmpC, _MM_SHUFFLE(3, 3, 2, 0)); 536 __m128 SwpFacC = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(2, 3, 3, 3)); 537 __m128 MulFacC = _mm_mul_ps(SwpFacC, SubFacC); 538 539 __m128 AddRes = _mm_add_ps(SubRes, MulFacC); 540 __m128 DetCof = _mm_mul_ps(AddRes, _mm_setr_ps( 1.0f,-1.0f, 1.0f,-1.0f)); 541 542 //return m[0][0] * DetCof[0] 543 // + m[0][1] * DetCof[1] 544 // + m[0][2] * DetCof[2] 545 // + m[0][3] * DetCof[3]; 546 547 return sse_dot_ps(m[0], DetCof); 548 } 549 550 GLM_FUNC_QUALIFIER void sse_inverse_ps(__m128 const in[4], __m128 out[4]) 551 { 552 __m128 Fac0; 553 { 554 // valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3]; 555 // valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3]; 556 // valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3]; 557 // valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3]; 558 559 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 560 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 561 562 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 563 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 564 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 565 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 566 567 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 568 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 569 Fac0 = _mm_sub_ps(Mul00, Mul01); 570 } 571 572 __m128 Fac1; 573 { 574 // valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3]; 575 // valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3]; 576 // valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3]; 577 // valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3]; 578 579 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 580 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 581 582 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 583 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 584 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 585 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 586 587 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 588 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 589 Fac1 = _mm_sub_ps(Mul00, Mul01); 590 } 591 592 593 __m128 Fac2; 594 { 595 // valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2]; 596 // valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2]; 597 // valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2]; 598 // valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2]; 599 600 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 601 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 602 603 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 604 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 605 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 606 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 607 608 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 609 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 610 Fac2 = _mm_sub_ps(Mul00, Mul01); 611 } 612 613 __m128 Fac3; 614 { 615 // valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3]; 616 // valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3]; 617 // valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3]; 618 // valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3]; 619 620 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 621 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 622 623 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 624 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 625 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 626 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 627 628 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 629 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 630 Fac3 = _mm_sub_ps(Mul00, Mul01); 631 } 632 633 __m128 Fac4; 634 { 635 // valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2]; 636 // valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2]; 637 // valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2]; 638 // valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2]; 639 640 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 641 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 642 643 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 644 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 645 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 646 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 647 648 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 649 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 650 Fac4 = _mm_sub_ps(Mul00, Mul01); 651 } 652 653 __m128 Fac5; 654 { 655 // valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1]; 656 // valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1]; 657 // valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1]; 658 // valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1]; 659 660 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 661 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 662 663 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 664 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 665 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 666 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 667 668 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 669 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 670 Fac5 = _mm_sub_ps(Mul00, Mul01); 671 } 672 673 __m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f); 674 __m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f); 675 676 // m[1][0] 677 // m[0][0] 678 // m[0][0] 679 // m[0][0] 680 __m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0)); 681 __m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0)); 682 683 // m[1][1] 684 // m[0][1] 685 // m[0][1] 686 // m[0][1] 687 __m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1)); 688 __m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0)); 689 690 // m[1][2] 691 // m[0][2] 692 // m[0][2] 693 // m[0][2] 694 __m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2)); 695 __m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0)); 696 697 // m[1][3] 698 // m[0][3] 699 // m[0][3] 700 // m[0][3] 701 __m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3)); 702 __m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0)); 703 704 // col0 705 // + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]), 706 // - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]), 707 // + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]), 708 // - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]), 709 __m128 Mul00 = _mm_mul_ps(Vec1, Fac0); 710 __m128 Mul01 = _mm_mul_ps(Vec2, Fac1); 711 __m128 Mul02 = _mm_mul_ps(Vec3, Fac2); 712 __m128 Sub00 = _mm_sub_ps(Mul00, Mul01); 713 __m128 Add00 = _mm_add_ps(Sub00, Mul02); 714 __m128 Inv0 = _mm_mul_ps(SignB, Add00); 715 716 // col1 717 // - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]), 718 // + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]), 719 // - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]), 720 // + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]), 721 __m128 Mul03 = _mm_mul_ps(Vec0, Fac0); 722 __m128 Mul04 = _mm_mul_ps(Vec2, Fac3); 723 __m128 Mul05 = _mm_mul_ps(Vec3, Fac4); 724 __m128 Sub01 = _mm_sub_ps(Mul03, Mul04); 725 __m128 Add01 = _mm_add_ps(Sub01, Mul05); 726 __m128 Inv1 = _mm_mul_ps(SignA, Add01); 727 728 // col2 729 // + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]), 730 // - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]), 731 // + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]), 732 // - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]), 733 __m128 Mul06 = _mm_mul_ps(Vec0, Fac1); 734 __m128 Mul07 = _mm_mul_ps(Vec1, Fac3); 735 __m128 Mul08 = _mm_mul_ps(Vec3, Fac5); 736 __m128 Sub02 = _mm_sub_ps(Mul06, Mul07); 737 __m128 Add02 = _mm_add_ps(Sub02, Mul08); 738 __m128 Inv2 = _mm_mul_ps(SignB, Add02); 739 740 // col3 741 // - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]), 742 // + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]), 743 // - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]), 744 // + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3])); 745 __m128 Mul09 = _mm_mul_ps(Vec0, Fac2); 746 __m128 Mul10 = _mm_mul_ps(Vec1, Fac4); 747 __m128 Mul11 = _mm_mul_ps(Vec2, Fac5); 748 __m128 Sub03 = _mm_sub_ps(Mul09, Mul10); 749 __m128 Add03 = _mm_add_ps(Sub03, Mul11); 750 __m128 Inv3 = _mm_mul_ps(SignA, Add03); 751 752 __m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0)); 753 __m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0)); 754 __m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0)); 755 756 // valType Determinant = m[0][0] * Inverse[0][0] 757 // + m[0][1] * Inverse[1][0] 758 // + m[0][2] * Inverse[2][0] 759 // + m[0][3] * Inverse[3][0]; 760 __m128 Det0 = sse_dot_ps(in[0], Row2); 761 __m128 Rcp0 = _mm_div_ps(one, Det0); 762 //__m128 Rcp0 = _mm_rcp_ps(Det0); 763 764 // Inverse /= Determinant; 765 out[0] = _mm_mul_ps(Inv0, Rcp0); 766 out[1] = _mm_mul_ps(Inv1, Rcp0); 767 out[2] = _mm_mul_ps(Inv2, Rcp0); 768 out[3] = _mm_mul_ps(Inv3, Rcp0); 769 } 770 771 GLM_FUNC_QUALIFIER void sse_inverse_fast_ps(__m128 const in[4], __m128 out[4]) 772 { 773 __m128 Fac0; 774 { 775 // valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3]; 776 // valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3]; 777 // valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3]; 778 // valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3]; 779 780 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 781 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 782 783 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 784 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 785 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 786 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 787 788 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 789 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 790 Fac0 = _mm_sub_ps(Mul00, Mul01); 791 } 792 793 __m128 Fac1; 794 { 795 // valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3]; 796 // valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3]; 797 // valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3]; 798 // valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3]; 799 800 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 801 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 802 803 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 804 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 805 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 806 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 807 808 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 809 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 810 Fac1 = _mm_sub_ps(Mul00, Mul01); 811 } 812 813 814 __m128 Fac2; 815 { 816 // valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2]; 817 // valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2]; 818 // valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2]; 819 // valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2]; 820 821 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 822 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 823 824 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 825 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 826 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 827 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 828 829 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 830 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 831 Fac2 = _mm_sub_ps(Mul00, Mul01); 832 } 833 834 __m128 Fac3; 835 { 836 // valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3]; 837 // valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3]; 838 // valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3]; 839 // valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3]; 840 841 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3)); 842 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 843 844 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 845 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 846 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 847 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3)); 848 849 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 850 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 851 Fac3 = _mm_sub_ps(Mul00, Mul01); 852 } 853 854 __m128 Fac4; 855 { 856 // valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2]; 857 // valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2]; 858 // valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2]; 859 // valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2]; 860 861 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2)); 862 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 863 864 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 865 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 866 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 867 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2)); 868 869 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 870 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 871 Fac4 = _mm_sub_ps(Mul00, Mul01); 872 } 873 874 __m128 Fac5; 875 { 876 // valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1]; 877 // valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1]; 878 // valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1]; 879 // valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1]; 880 881 __m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1)); 882 __m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0)); 883 884 __m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0)); 885 __m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0)); 886 __m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0)); 887 __m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1)); 888 889 __m128 Mul00 = _mm_mul_ps(Swp00, Swp01); 890 __m128 Mul01 = _mm_mul_ps(Swp02, Swp03); 891 Fac5 = _mm_sub_ps(Mul00, Mul01); 892 } 893 894 __m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f); 895 __m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f); 896 897 // m[1][0] 898 // m[0][0] 899 // m[0][0] 900 // m[0][0] 901 __m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0)); 902 __m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0)); 903 904 // m[1][1] 905 // m[0][1] 906 // m[0][1] 907 // m[0][1] 908 __m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1)); 909 __m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0)); 910 911 // m[1][2] 912 // m[0][2] 913 // m[0][2] 914 // m[0][2] 915 __m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2)); 916 __m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0)); 917 918 // m[1][3] 919 // m[0][3] 920 // m[0][3] 921 // m[0][3] 922 __m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3)); 923 __m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0)); 924 925 // col0 926 // + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]), 927 // - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]), 928 // + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]), 929 // - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]), 930 __m128 Mul00 = _mm_mul_ps(Vec1, Fac0); 931 __m128 Mul01 = _mm_mul_ps(Vec2, Fac1); 932 __m128 Mul02 = _mm_mul_ps(Vec3, Fac2); 933 __m128 Sub00 = _mm_sub_ps(Mul00, Mul01); 934 __m128 Add00 = _mm_add_ps(Sub00, Mul02); 935 __m128 Inv0 = _mm_mul_ps(SignB, Add00); 936 937 // col1 938 // - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]), 939 // + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]), 940 // - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]), 941 // + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]), 942 __m128 Mul03 = _mm_mul_ps(Vec0, Fac0); 943 __m128 Mul04 = _mm_mul_ps(Vec2, Fac3); 944 __m128 Mul05 = _mm_mul_ps(Vec3, Fac4); 945 __m128 Sub01 = _mm_sub_ps(Mul03, Mul04); 946 __m128 Add01 = _mm_add_ps(Sub01, Mul05); 947 __m128 Inv1 = _mm_mul_ps(SignA, Add01); 948 949 // col2 950 // + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]), 951 // - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]), 952 // + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]), 953 // - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]), 954 __m128 Mul06 = _mm_mul_ps(Vec0, Fac1); 955 __m128 Mul07 = _mm_mul_ps(Vec1, Fac3); 956 __m128 Mul08 = _mm_mul_ps(Vec3, Fac5); 957 __m128 Sub02 = _mm_sub_ps(Mul06, Mul07); 958 __m128 Add02 = _mm_add_ps(Sub02, Mul08); 959 __m128 Inv2 = _mm_mul_ps(SignB, Add02); 960 961 // col3 962 // - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]), 963 // + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]), 964 // - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]), 965 // + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3])); 966 __m128 Mul09 = _mm_mul_ps(Vec0, Fac2); 967 __m128 Mul10 = _mm_mul_ps(Vec1, Fac4); 968 __m128 Mul11 = _mm_mul_ps(Vec2, Fac5); 969 __m128 Sub03 = _mm_sub_ps(Mul09, Mul10); 970 __m128 Add03 = _mm_add_ps(Sub03, Mul11); 971 __m128 Inv3 = _mm_mul_ps(SignA, Add03); 972 973 __m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0)); 974 __m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0)); 975 __m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0)); 976 977 // valType Determinant = m[0][0] * Inverse[0][0] 978 // + m[0][1] * Inverse[1][0] 979 // + m[0][2] * Inverse[2][0] 980 // + m[0][3] * Inverse[3][0]; 981 __m128 Det0 = sse_dot_ps(in[0], Row2); 982 __m128 Rcp0 = _mm_rcp_ps(Det0); 983 //__m128 Rcp0 = _mm_div_ps(one, Det0); 984 // Inverse /= Determinant; 985 out[0] = _mm_mul_ps(Inv0, Rcp0); 986 out[1] = _mm_mul_ps(Inv1, Rcp0); 987 out[2] = _mm_mul_ps(Inv2, Rcp0); 988 out[3] = _mm_mul_ps(Inv3, Rcp0); 989 } 990 /* 991 GLM_FUNC_QUALIFIER void sse_rotate_ps(__m128 const in[4], float Angle, float const v[3], __m128 out[4]) 992 { 993 float a = glm::radians(Angle); 994 float c = cos(a); 995 float s = sin(a); 996 997 glm::vec4 AxisA(v[0], v[1], v[2], float(0)); 998 __m128 AxisB = _mm_set_ps(AxisA.w, AxisA.z, AxisA.y, AxisA.x); 999 __m128 AxisC = detail::sse_nrm_ps(AxisB); 1000 1001 __m128 Cos0 = _mm_set_ss(c); 1002 __m128 CosA = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(0, 0, 0, 0)); 1003 __m128 Sin0 = _mm_set_ss(s); 1004 __m128 SinA = _mm_shuffle_ps(Sin0, Sin0, _MM_SHUFFLE(0, 0, 0, 0)); 1005 1006 // detail::tvec3<T, P> temp = (valType(1) - c) * axis; 1007 __m128 Temp0 = _mm_sub_ps(one, CosA); 1008 __m128 Temp1 = _mm_mul_ps(Temp0, AxisC); 1009 1010 //Rotate[0][0] = c + temp[0] * axis[0]; 1011 //Rotate[0][1] = 0 + temp[0] * axis[1] + s * axis[2]; 1012 //Rotate[0][2] = 0 + temp[0] * axis[2] - s * axis[1]; 1013 __m128 Axis0 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(0, 0, 0, 0)); 1014 __m128 TmpA0 = _mm_mul_ps(Axis0, AxisC); 1015 __m128 CosA0 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 1, 1, 0)); 1016 __m128 TmpA1 = _mm_add_ps(CosA0, TmpA0); 1017 __m128 SinA0 = SinA;//_mm_set_ps(0.0f, s, -s, 0.0f); 1018 __m128 TmpA2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 1, 2, 3)); 1019 __m128 TmpA3 = _mm_mul_ps(SinA0, TmpA2); 1020 __m128 TmpA4 = _mm_add_ps(TmpA1, TmpA3); 1021 1022 //Rotate[1][0] = 0 + temp[1] * axis[0] - s * axis[2]; 1023 //Rotate[1][1] = c + temp[1] * axis[1]; 1024 //Rotate[1][2] = 0 + temp[1] * axis[2] + s * axis[0]; 1025 __m128 Axis1 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(1, 1, 1, 1)); 1026 __m128 TmpB0 = _mm_mul_ps(Axis1, AxisC); 1027 __m128 CosA1 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 1, 0, 1)); 1028 __m128 TmpB1 = _mm_add_ps(CosA1, TmpB0); 1029 __m128 SinB0 = SinA;//_mm_set_ps(-s, 0.0f, s, 0.0f); 1030 __m128 TmpB2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 0, 3, 2)); 1031 __m128 TmpB3 = _mm_mul_ps(SinA0, TmpB2); 1032 __m128 TmpB4 = _mm_add_ps(TmpB1, TmpB3); 1033 1034 //Rotate[2][0] = 0 + temp[2] * axis[0] + s * axis[1]; 1035 //Rotate[2][1] = 0 + temp[2] * axis[1] - s * axis[0]; 1036 //Rotate[2][2] = c + temp[2] * axis[2]; 1037 __m128 Axis2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(2, 2, 2, 2)); 1038 __m128 TmpC0 = _mm_mul_ps(Axis2, AxisC); 1039 __m128 CosA2 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 0, 1, 1)); 1040 __m128 TmpC1 = _mm_add_ps(CosA2, TmpC0); 1041 __m128 SinC0 = SinA;//_mm_set_ps(s, -s, 0.0f, 0.0f); 1042 __m128 TmpC2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 3, 0, 1)); 1043 __m128 TmpC3 = _mm_mul_ps(SinA0, TmpC2); 1044 __m128 TmpC4 = _mm_add_ps(TmpC1, TmpC3); 1045 1046 __m128 Result[4]; 1047 Result[0] = TmpA4; 1048 Result[1] = TmpB4; 1049 Result[2] = TmpC4; 1050 Result[3] = _mm_set_ps(1, 0, 0, 0); 1051 1052 //detail::tmat4x4<valType> Result(detail::tmat4x4<valType>::_null); 1053 //Result[0] = m[0] * Rotate[0][0] + m[1] * Rotate[0][1] + m[2] * Rotate[0][2]; 1054 //Result[1] = m[0] * Rotate[1][0] + m[1] * Rotate[1][1] + m[2] * Rotate[1][2]; 1055 //Result[2] = m[0] * Rotate[2][0] + m[1] * Rotate[2][1] + m[2] * Rotate[2][2]; 1056 //Result[3] = m[3]; 1057 //return Result; 1058 sse_mul_ps(in, Result, out); 1059 } 1060 */ 1061 GLM_FUNC_QUALIFIER void sse_outer_ps(__m128 const & c, __m128 const & r, __m128 out[4]) 1062 { 1063 out[0] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(0, 0, 0, 0))); 1064 out[1] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(1, 1, 1, 1))); 1065 out[2] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(2, 2, 2, 2))); 1066 out[3] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(3, 3, 3, 3))); 1067 } 1068 1069 }//namespace detail 1070 }//namespace glm 1071
