constexpr T _pdep(T src, T mask)
{
static_assert(binary_digits<T>::value, "");
constexpr T digits = binary_digits<T>::value;
T dest = 0;
if (digits <= std::numeric_limits<unsigned int>::digits) {
dest = _pdep_u32(src, mask);
} else if (digits <= std::numeric_limits<unsigned long long int>::digits) {
dest = _pdep_u64(src, mask);
} else {
dest = _pdep(src, mask, std::ignore);
}
return dest;
}
unsigned long long test_pdep_u64(unsigned long long __X, unsigned long long __Y) {
// CHECK: @llvm.x86.bmi.pdep.64
return _pdep_u64(__X, __Y);
}
constexpr T _lzcnt(T src, X...); // Trailing zeros count template <class T, class = decltype(__builtin_ctzll(T()))> constexpr T _tzcnt(T src); template <class T, class... X> constexpr T _tzcnt(T src, X...); // Bit field extraction template <class T, class = decltype(__builtin_ia32_bextr_u64(T(), T(), T()))> constexpr T _bextr(T src, T start, T len); template <class T, class... X> constexpr T _bextr(T src, T start, T len, X...); // Parallel bits deposit template <class T, class = decltype(_pdep_u64(T()))> constexpr T _pdep(T src, T mask); template <class T, class... X> constexpr T _pdep(T src, T mask, X...); // Parallel bits extract template <class T, class = decltype(_pext_u64(T()))> constexpr T _pext(T src, T mask); template <class T, class... X> constexpr T _pext(T src, T mask, X...); // Byte swap template <class T, class = decltype(__builtin_bswap64(T()))> constexpr T _byteswap(T src); template <class T, class... X> constexpr T _byteswap(T src, X...);
uint64_t linearize(uint32_t x, uint32_t y) {
uint64_t result;
result = _pdep_u64(x, 0x5555555555555555);
result ^= _pdep_u64(y, 0xAAAAAAAAAAAAAAAA);
return result;
}