Function core::arch::x86_64::_mm_addsub_ps

pub unsafe fn _mm_addsub_ps(a: __m128, b: __m128) -> __m128

This is supported on x86-64 and target feature sse3 only.

Alternatively add and subtract packed single-precision (32-bit) floating-point elements in a to/from packed elements in b.

Intel's documentation