}

template<classA,classT,bool... Values,classMode,class =typename std::enable_if<std::is_integral<T>::value>::type>

XSIMD_INLINE batch<T, A>load_masked(Tconst* mem, batch_bool_constant<T, A, Values...> mask, Mode, requires_arch<sse2>)noexcept

{

XSIMD_IF_CONSTEXPR(mask.mask() ==0x1)

{

XSIMD_IF_CONSTEXPR(sizeof(T) ==2)

{

returnmm_loadu_si16(mem);

}

XSIMD_IF_CONSTEXPR(sizeof(T) ==4)

{

returnmm_loadu_si32(mem);

}

XSIMD_IF_CONSTEXPR(sizeof(T) ==8)

{

returnmm_loadu_si64(mem);

}

}

elseXSIMD_IF_CONSTEXPR(sizeof(T) ==2 && mask.mask() ==0x3)

{

returnmm_loadu_si32(mem);

}

elseXSIMD_IF_CONSTEXPR(sizeof(T) ==4 && mask.mask() ==0x3)

{

returnmm_loadu_si64(mem);

}

{

return load_masked<A>(mem, mask, convert<T> {}, Mode {}, common {});

}

}

template<classA,bool... Values,classMode>

XSIMD_INLINE batch<float, A>load_masked(floatconst* mem, batch_bool_constant<float, A, Values...> mask, Mode, requires_arch<sse2>)noexcept

{

XSIMD_IF_CONSTEXPR(mask.countr_one() ==2)

XSIMD_IF_CONSTEXPR(mask.mask() ==0x1)

{

return_mm_load_ss(mem);

}

elseXSIMD_IF_CONSTEXPR(mask.countr_one() ==2)

{

return_mm_loadl_pi(_mm_setzero_ps(),reinterpret_cast<__m64const*>(mem));

}

{

XSIMD_IF_CONSTEXPR(mask.countr_one() ==1)

{

return_mm_move_sd(_mm_setzero_pd(),_mm_load_sd(mem));

return_mm_load_sd(mem);

}

elseXSIMD_IF_CONSTEXPR(mask.countl_one() ==1)

{

template<classA,bool... Values,classMode>

XSIMD_INLINEvoidstore_masked(float* mem, batch<float, A>const& src, batch_bool_constant<float, A, Values...> mask, Mode, requires_arch<sse2>)noexcept

{

XSIMD_IF_CONSTEXPR(mask.countr_one() ==2)

XSIMD_IF_CONSTEXPR(mask.mask() ==0x1)

{

_mm_store_ss(mem, src);

}

elseXSIMD_IF_CONSTEXPR(mask.countr_one() ==2)

{

_mm_storel_pi(reinterpret_cast<__m64*>(mem), src);

}