Function core_arch::x86::_mm_madd_epi16

pub unsafe fn _mm_madd_epi16(a: __m128i, b: __m128i) -> __m128i

This is supported on x86 and target feature sse2 only.

Multiplies and then horizontally add signed 16 bit integers in a and b.

Multiplies packed signed 16-bit integers in a and b, producing intermediate signed 32-bit integers. Horizontally add adjacent pairs of intermediate 32-bit integers.

Intel's documentation