, , GCC. set/load/store. _mm_setr_ps , . , .
a, /
__m128 t = _mm_loadu_ps(a);
_mm_storeu_ps(a, t);
16 , /, , .
__m128 t = _mm_load_ps(a);
_mm_store_ps(a, t);
16- ,
__declspec(align(16)) const float a[] = ...
__attribute__((aligned(16))) const float a[] ...
16- :
float *a = (float*)_mm_malloc(sizeof(float)*n, 16);