Question

The code below has me slightly perplexed:

function(__m128 foo)
{
  __m128 bar = _mm_shuffle_ps(foo, foo, _MM_SHUFFLE(2,2,2,2))
}

Is it just taking the 2nd word of foo and pasting it 4 times into bar or does it do something else as well?

Answer 1

I don’t know if _mm_shuffle exists. It should rather be _mm_shuffle_ps, given the __m128 arguments.

In this case it does exactly what you think. It copies the 3rd (not the 2nd, counting starts at 0, from the right, so it’s indeed the 2nd from the left) 32-bit word of foo into all 4 32-bit words of bar. What else should it do?