pub unsafe fn _mm_hadd_ps(a: __m128, b: __m128) -> __m128
Horizontally adds adjacent pairs of single-precision (32-bit) floating-point elements in a and b, and pack the results.
a
b
Intel’s documentation