| 1 | /// @ref core |
| 2 | |
| 3 | #include "_vectorize.hpp" |
| 4 | #if(GLM_ARCH & GLM_ARCH_X86 && GLM_COMPILER & GLM_COMPILER_VC) |
| 5 | # include <intrin.h> |
| 6 | # pragma intrinsic(_BitScanReverse) |
| 7 | #endif//(GLM_ARCH & GLM_ARCH_X86 && GLM_COMPILER & GLM_COMPILER_VC) |
| 8 | #include <limits> |
| 9 | |
| 10 | #if !GLM_HAS_EXTENDED_INTEGER_TYPE |
| 11 | # if GLM_COMPILER & GLM_COMPILER_GCC |
| 12 | # pragma GCC diagnostic ignored "-Wlong-long" |
| 13 | # endif |
| 14 | # if (GLM_COMPILER & GLM_COMPILER_CLANG) |
| 15 | # pragma clang diagnostic ignored "-Wc++11-long-long" |
| 16 | # endif |
| 17 | #endif |
| 18 | |
| 19 | namespace glm{ |
| 20 | namespace detail |
| 21 | { |
| 22 | template<typename T> |
| 23 | GLM_FUNC_QUALIFIER T mask(T Bits) |
| 24 | { |
| 25 | return Bits >= static_cast<T>(sizeof(T) * 8) ? ~static_cast<T>(0) : (static_cast<T>(1) << Bits) - static_cast<T>(1); |
| 26 | } |
| 27 | |
| 28 | template<length_t L, typename T, qualifier Q, bool Aligned, bool EXEC> |
| 29 | struct compute_bitfieldReverseStep |
| 30 | { |
| 31 | GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T, T) |
| 32 | { |
| 33 | return v; |
| 34 | } |
| 35 | }; |
| 36 | |
| 37 | template<length_t L, typename T, qualifier Q, bool Aligned> |
| 38 | struct compute_bitfieldReverseStep<L, T, Q, Aligned, true> |
| 39 | { |
| 40 | GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T Mask, T Shift) |
| 41 | { |
| 42 | return (v & Mask) << Shift | (v & (~Mask)) >> Shift; |
| 43 | } |
| 44 | }; |
| 45 | |
| 46 | template<length_t L, typename T, qualifier Q, bool Aligned, bool EXEC> |
| 47 | struct compute_bitfieldBitCountStep |
| 48 | { |
| 49 | GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T, T) |
| 50 | { |
| 51 | return v; |
| 52 | } |
| 53 | }; |
| 54 | |
| 55 | template<length_t L, typename T, qualifier Q, bool Aligned> |
| 56 | struct compute_bitfieldBitCountStep<L, T, Q, Aligned, true> |
| 57 | { |
| 58 | GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& v, T Mask, T Shift) |
| 59 | { |
| 60 | return (v & Mask) + ((v >> Shift) & Mask); |
| 61 | } |
| 62 | }; |
| 63 | |
| 64 | template<typename genIUType, size_t Bits> |
| 65 | struct compute_findLSB |
| 66 | { |
| 67 | GLM_FUNC_QUALIFIER static int call(genIUType Value) |
| 68 | { |
| 69 | if(Value == 0) |
| 70 | return -1; |
| 71 | |
| 72 | return glm::bitCount(~Value & (Value - static_cast<genIUType>(1))); |
| 73 | } |
| 74 | }; |
| 75 | |
| 76 | # if GLM_HAS_BITSCAN_WINDOWS |
| 77 | template<typename genIUType> |
| 78 | struct compute_findLSB<genIUType, 32> |
| 79 | { |
| 80 | GLM_FUNC_QUALIFIER static int call(genIUType Value) |
| 81 | { |
| 82 | unsigned long Result(0); |
| 83 | unsigned char IsNotNull = _BitScanForward(&Result, *reinterpret_cast<unsigned long*>(&Value)); |
| 84 | return IsNotNull ? int(Result) : -1; |
| 85 | } |
| 86 | }; |
| 87 | |
| 88 | # if !((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_MODEL == GLM_MODEL_32)) |
| 89 | template<typename genIUType> |
| 90 | struct compute_findLSB<genIUType, 64> |
| 91 | { |
| 92 | GLM_FUNC_QUALIFIER static int call(genIUType Value) |
| 93 | { |
| 94 | unsigned long Result(0); |
| 95 | unsigned char IsNotNull = _BitScanForward64(&Result, *reinterpret_cast<unsigned __int64*>(&Value)); |
| 96 | return IsNotNull ? int(Result) : -1; |
| 97 | } |
| 98 | }; |
| 99 | # endif |
| 100 | # endif//GLM_HAS_BITSCAN_WINDOWS |
| 101 | |
| 102 | template<length_t L, typename T, qualifier Q, bool EXEC = true> |
| 103 | struct compute_findMSB_step_vec |
| 104 | { |
| 105 | GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, T Shift) |
| 106 | { |
| 107 | return x | (x >> Shift); |
| 108 | } |
| 109 | }; |
| 110 | |
| 111 | template<length_t L, typename T, qualifier Q> |
| 112 | struct compute_findMSB_step_vec<L, T, Q, false> |
| 113 | { |
| 114 | GLM_FUNC_QUALIFIER static vec<L, T, Q> call(vec<L, T, Q> const& x, T) |
| 115 | { |
| 116 | return x; |
| 117 | } |
| 118 | }; |
| 119 | |
| 120 | template<length_t L, typename T, qualifier Q, int> |
| 121 | struct compute_findMSB_vec |
| 122 | { |
| 123 | GLM_FUNC_QUALIFIER static vec<L, int, Q> call(vec<L, T, Q> const& v) |
| 124 | { |
| 125 | vec<L, T, Q> x(v); |
| 126 | x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 8>::call(x, static_cast<T>( 1)); |
| 127 | x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 8>::call(x, static_cast<T>( 2)); |
| 128 | x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 8>::call(x, static_cast<T>( 4)); |
| 129 | x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 16>::call(x, static_cast<T>( 8)); |
| 130 | x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 32>::call(x, static_cast<T>(16)); |
| 131 | x = compute_findMSB_step_vec<L, T, Q, sizeof(T) * 8 >= 64>::call(x, static_cast<T>(32)); |
| 132 | return vec<L, int, Q>(sizeof(T) * 8 - 1) - glm::bitCount(~x); |
| 133 | } |
| 134 | }; |
| 135 | |
| 136 | # if GLM_HAS_BITSCAN_WINDOWS |
| 137 | template<typename genIUType> |
| 138 | GLM_FUNC_QUALIFIER int compute_findMSB_32(genIUType Value) |
| 139 | { |
| 140 | unsigned long Result(0); |
| 141 | unsigned char IsNotNull = _BitScanReverse(&Result, *reinterpret_cast<unsigned long*>(&Value)); |
| 142 | return IsNotNull ? int(Result) : -1; |
| 143 | } |
| 144 | |
| 145 | template<length_t L, typename T, qualifier Q> |
| 146 | struct compute_findMSB_vec<L, T, Q, 32> |
| 147 | { |
| 148 | GLM_FUNC_QUALIFIER static vec<L, int, Q> call(vec<L, T, Q> const& x) |
| 149 | { |
| 150 | return detail::functor1<vec, L, int, T, Q>::call(compute_findMSB_32, x); |
| 151 | } |
| 152 | }; |
| 153 | |
| 154 | # if !((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_MODEL == GLM_MODEL_32)) |
| 155 | template<typename genIUType> |
| 156 | GLM_FUNC_QUALIFIER int compute_findMSB_64(genIUType Value) |
| 157 | { |
| 158 | unsigned long Result(0); |
| 159 | unsigned char IsNotNull = _BitScanReverse64(&Result, *reinterpret_cast<unsigned __int64*>(&Value)); |
| 160 | return IsNotNull ? int(Result) : -1; |
| 161 | } |
| 162 | |
| 163 | template<length_t L, typename T, qualifier Q> |
| 164 | struct compute_findMSB_vec<L, T, Q, 64> |
| 165 | { |
| 166 | GLM_FUNC_QUALIFIER static vec<L, int, Q> call(vec<L, T, Q> const& x) |
| 167 | { |
| 168 | return detail::functor1<vec, L, int, T, Q>::call(compute_findMSB_64, x); |
| 169 | } |
| 170 | }; |
| 171 | # endif |
| 172 | # endif//GLM_HAS_BITSCAN_WINDOWS |
| 173 | }//namespace detail |
| 174 | |
| 175 | // uaddCarry |
| 176 | GLM_FUNC_QUALIFIER uint uaddCarry(uint const& x, uint const& y, uint & Carry) |
| 177 | { |
| 178 | detail::uint64 const Value64(static_cast<detail::uint64>(x) + static_cast<detail::uint64>(y)); |
| 179 | detail::uint64 const Max32((static_cast<detail::uint64>(1) << static_cast<detail::uint64>(32)) - static_cast<detail::uint64>(1)); |
| 180 | Carry = Value64 > Max32 ? 1u : 0u; |
| 181 | return static_cast<uint>(Value64 % (Max32 + static_cast<detail::uint64>(1))); |
| 182 | } |
| 183 | |
| 184 | template<length_t L, qualifier Q> |
| 185 | GLM_FUNC_QUALIFIER vec<L, uint, Q> uaddCarry(vec<L, uint, Q> const& x, vec<L, uint, Q> const& y, vec<L, uint, Q>& Carry) |
| 186 | { |
| 187 | vec<L, detail::uint64, Q> Value64(vec<L, detail::uint64, Q>(x) + vec<L, detail::uint64, Q>(y)); |
| 188 | vec<L, detail::uint64, Q> Max32((static_cast<detail::uint64>(1) << static_cast<detail::uint64>(32)) - static_cast<detail::uint64>(1)); |
| 189 | Carry = mix(vec<L, uint, Q>(0), vec<L, uint, Q>(1), greaterThan(Value64, Max32)); |
| 190 | return vec<L, uint, Q>(Value64 % (Max32 + static_cast<detail::uint64>(1))); |
| 191 | } |
| 192 | |
| 193 | // usubBorrow |
| 194 | GLM_FUNC_QUALIFIER uint usubBorrow(uint const& x, uint const& y, uint & Borrow) |
| 195 | { |
| 196 | Borrow = x >= y ? static_cast<uint>(0) : static_cast<uint>(1); |
| 197 | if(y >= x) |
| 198 | return y - x; |
| 199 | else |
| 200 | return static_cast<uint>((static_cast<detail::int64>(1) << static_cast<detail::int64>(32)) + (static_cast<detail::int64>(y) - static_cast<detail::int64>(x))); |
| 201 | } |
| 202 | |
| 203 | template<length_t L, qualifier Q> |
| 204 | GLM_FUNC_QUALIFIER vec<L, uint, Q> usubBorrow(vec<L, uint, Q> const& x, vec<L, uint, Q> const& y, vec<L, uint, Q>& Borrow) |
| 205 | { |
| 206 | Borrow = mix(vec<L, uint, Q>(1), vec<L, uint, Q>(0), greaterThanEqual(x, y)); |
| 207 | vec<L, uint, Q> const YgeX(y - x); |
| 208 | vec<L, uint, Q> const XgeY(vec<L, uint, Q>((static_cast<detail::int64>(1) << static_cast<detail::int64>(32)) + (vec<L, detail::int64, Q>(y) - vec<L, detail::int64, Q>(x)))); |
| 209 | return mix(XgeY, YgeX, greaterThanEqual(y, x)); |
| 210 | } |
| 211 | |
| 212 | // umulExtended |
| 213 | GLM_FUNC_QUALIFIER void umulExtended(uint const& x, uint const& y, uint & msb, uint & lsb) |
| 214 | { |
| 215 | detail::uint64 Value64 = static_cast<detail::uint64>(x) * static_cast<detail::uint64>(y); |
| 216 | msb = static_cast<uint>(Value64 >> static_cast<detail::uint64>(32)); |
| 217 | lsb = static_cast<uint>(Value64); |
| 218 | } |
| 219 | |
| 220 | template<length_t L, qualifier Q> |
| 221 | GLM_FUNC_QUALIFIER void umulExtended(vec<L, uint, Q> const& x, vec<L, uint, Q> const& y, vec<L, uint, Q>& msb, vec<L, uint, Q>& lsb) |
| 222 | { |
| 223 | vec<L, detail::uint64, Q> Value64(vec<L, detail::uint64, Q>(x) * vec<L, detail::uint64, Q>(y)); |
| 224 | msb = vec<L, uint, Q>(Value64 >> static_cast<detail::uint64>(32)); |
| 225 | lsb = vec<L, uint, Q>(Value64); |
| 226 | } |
| 227 | |
| 228 | // imulExtended |
| 229 | GLM_FUNC_QUALIFIER void imulExtended(int x, int y, int& msb, int& lsb) |
| 230 | { |
| 231 | detail::int64 Value64 = static_cast<detail::int64>(x) * static_cast<detail::int64>(y); |
| 232 | msb = static_cast<int>(Value64 >> static_cast<detail::int64>(32)); |
| 233 | lsb = static_cast<int>(Value64); |
| 234 | } |
| 235 | |
| 236 | template<length_t L, qualifier Q> |
| 237 | GLM_FUNC_QUALIFIER void imulExtended(vec<L, int, Q> const& x, vec<L, int, Q> const& y, vec<L, int, Q>& msb, vec<L, int, Q>& lsb) |
| 238 | { |
| 239 | vec<L, detail::int64, Q> Value64(vec<L, detail::int64, Q>(x) * vec<L, detail::int64, Q>(y)); |
| 240 | lsb = vec<L, int, Q>(Value64 & static_cast<detail::int64>(0xFFFFFFFF)); |
| 241 | msb = vec<L, int, Q>((Value64 >> static_cast<detail::int64>(32)) & static_cast<detail::int64>(0xFFFFFFFF)); |
| 242 | } |
| 243 | |
| 244 | // bitfieldExtract |
| 245 | template<typename genIUType> |
| 246 | GLM_FUNC_QUALIFIER genIUType bitfieldExtract(genIUType Value, int Offset, int Bits) |
| 247 | { |
| 248 | return bitfieldExtract(vec<1, genIUType>(Value), Offset, Bits).x; |
| 249 | } |
| 250 | |
| 251 | template<length_t L, typename T, qualifier Q> |
| 252 | GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldExtract(vec<L, T, Q> const& Value, int Offset, int Bits) |
| 253 | { |
| 254 | GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldExtract' only accept integer inputs"); |
| 255 | |
| 256 | return (Value >> static_cast<T>(Offset)) & static_cast<T>(detail::mask(Bits)); |
| 257 | } |
| 258 | |
| 259 | // bitfieldInsert |
| 260 | template<typename genIUType> |
| 261 | GLM_FUNC_QUALIFIER genIUType bitfieldInsert(genIUType const& Base, genIUType const& Insert, int Offset, int Bits) |
| 262 | { |
| 263 | return bitfieldInsert(vec<1, genIUType>(Base), vec<1, genIUType>(Insert), Offset, Bits).x; |
| 264 | } |
| 265 | |
| 266 | template<length_t L, typename T, qualifier Q> |
| 267 | GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldInsert(vec<L, T, Q> const& Base, vec<L, T, Q> const& Insert, int Offset, int Bits) |
| 268 | { |
| 269 | GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'bitfieldInsert' only accept integer values"); |
| 270 | |
| 271 | T const Mask = static_cast<T>(detail::mask(Bits) << Offset); |
| 272 | return (Base & ~Mask) | (Insert & Mask); |
| 273 | } |
| 274 | |
| 275 | // bitfieldReverse |
| 276 | template<typename genType> |
| 277 | GLM_FUNC_QUALIFIER genType bitfieldReverse(genType x) |
| 278 | { |
| 279 | return bitfieldReverse(glm::vec<1, genType, glm::defaultp>(x)).x; |
| 280 | } |
| 281 | |
| 282 | template<length_t L, typename T, qualifier Q> |
| 283 | GLM_FUNC_QUALIFIER vec<L, T, Q> bitfieldReverse(vec<L, T, Q> const& v) |
| 284 | { |
| 285 | vec<L, T, Q> x(v); |
| 286 | x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 2>::call(x, static_cast<T>(0x5555555555555555ull), static_cast<T>( 1)); |
| 287 | x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 4>::call(x, static_cast<T>(0x3333333333333333ull), static_cast<T>( 2)); |
| 288 | x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 8>::call(x, static_cast<T>(0x0F0F0F0F0F0F0F0Full), static_cast<T>( 4)); |
| 289 | x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 16>::call(x, static_cast<T>(0x00FF00FF00FF00FFull), static_cast<T>( 8)); |
| 290 | x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 32>::call(x, static_cast<T>(0x0000FFFF0000FFFFull), static_cast<T>(16)); |
| 291 | x = detail::compute_bitfieldReverseStep<L, T, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 64>::call(x, static_cast<T>(0x00000000FFFFFFFFull), static_cast<T>(32)); |
| 292 | return x; |
| 293 | } |
| 294 | |
| 295 | // bitCount |
| 296 | template<typename genType> |
| 297 | GLM_FUNC_QUALIFIER int bitCount(genType x) |
| 298 | { |
| 299 | return bitCount(glm::vec<1, genType, glm::defaultp>(x)).x; |
| 300 | } |
| 301 | |
| 302 | template<length_t L, typename T, qualifier Q> |
| 303 | GLM_FUNC_QUALIFIER vec<L, int, Q> bitCount(vec<L, T, Q> const& v) |
| 304 | { |
| 305 | # if GLM_COMPILER & GLM_COMPILER_VC |
| 306 | # pragma warning(push) |
| 307 | # pragma warning(disable : 4310) //cast truncates constant value |
| 308 | # endif |
| 309 | |
| 310 | vec<L, typename detail::make_unsigned<T>::type, Q> x(*reinterpret_cast<vec<L, typename detail::make_unsigned<T>::type, Q> const *>(&v)); |
| 311 | x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 2>::call(x, typename detail::make_unsigned<T>::type(0x5555555555555555ull), typename detail::make_unsigned<T>::type( 1)); |
| 312 | x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 4>::call(x, typename detail::make_unsigned<T>::type(0x3333333333333333ull), typename detail::make_unsigned<T>::type( 2)); |
| 313 | x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 8>::call(x, typename detail::make_unsigned<T>::type(0x0F0F0F0F0F0F0F0Full), typename detail::make_unsigned<T>::type( 4)); |
| 314 | x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 16>::call(x, typename detail::make_unsigned<T>::type(0x00FF00FF00FF00FFull), typename detail::make_unsigned<T>::type( 8)); |
| 315 | x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 32>::call(x, typename detail::make_unsigned<T>::type(0x0000FFFF0000FFFFull), typename detail::make_unsigned<T>::type(16)); |
| 316 | x = detail::compute_bitfieldBitCountStep<L, typename detail::make_unsigned<T>::type, Q, detail::is_aligned<Q>::value, sizeof(T) * 8>= 64>::call(x, typename detail::make_unsigned<T>::type(0x00000000FFFFFFFFull), typename detail::make_unsigned<T>::type(32)); |
| 317 | return vec<L, int, Q>(x); |
| 318 | |
| 319 | # if GLM_COMPILER & GLM_COMPILER_VC |
| 320 | # pragma warning(pop) |
| 321 | # endif |
| 322 | } |
| 323 | |
| 324 | // findLSB |
| 325 | template<typename genIUType> |
| 326 | GLM_FUNC_QUALIFIER int findLSB(genIUType Value) |
| 327 | { |
| 328 | GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'findLSB' only accept integer values"); |
| 329 | |
| 330 | return detail::compute_findLSB<genIUType, sizeof(genIUType) * 8>::call(Value); |
| 331 | } |
| 332 | |
| 333 | template<length_t L, typename T, qualifier Q> |
| 334 | GLM_FUNC_QUALIFIER vec<L, int, Q> findLSB(vec<L, T, Q> const& x) |
| 335 | { |
| 336 | GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'findLSB' only accept integer values"); |
| 337 | |
| 338 | return detail::functor1<vec, L, int, T, Q>::call(findLSB, x); |
| 339 | } |
| 340 | |
| 341 | // findMSB |
| 342 | template<typename genIUType> |
| 343 | GLM_FUNC_QUALIFIER int findMSB(genIUType v) |
| 344 | { |
| 345 | GLM_STATIC_ASSERT(std::numeric_limits<genIUType>::is_integer, "'findMSB' only accept integer values"); |
| 346 | |
| 347 | return findMSB(vec<1, genIUType>(v)).x; |
| 348 | } |
| 349 | |
| 350 | template<length_t L, typename T, qualifier Q> |
| 351 | GLM_FUNC_QUALIFIER vec<L, int, Q> findMSB(vec<L, T, Q> const& v) |
| 352 | { |
| 353 | GLM_STATIC_ASSERT(std::numeric_limits<T>::is_integer, "'findMSB' only accept integer values"); |
| 354 | |
| 355 | return detail::compute_findMSB_vec<L, T, Q, sizeof(T) * 8>::call(v); |
| 356 | } |
| 357 | }//namespace glm |
| 358 | |
| 359 | #if GLM_CONFIG_SIMD == GLM_ENABLE |
| 360 | # include "func_integer_simd.inl" |
| 361 | #endif |
| 362 | |