Question

How can I efficiently dump xmm register into uint8_t array[16]?

Answer 1

_mm_store_si128 stores the sse2 register contents to the memory specified.
_mm_store_ps and _mm_store_pd are the variants used for float or double register contents. When the destination is unaligned, you must use storeu.

__m128i var;
__declspec( align(16) ) uint8_t array[16];

_mm_store_si128( (__m128i*) array, var );