How does _mm256_shuffle_ps work?

Question

I understand how _mm_shuffle_ps work. For example, in the following.

__m128 r = _mm_shuffle_ps(x,y, _MM_SHUFFLE(2,0,2,0)); 

r will have contents, x[0], x[2], y[0], y[2].

But I see that _MM_SHUFFLE also takes 4 parameters for _mm256_shuffle_ps, while there the vectors would have 8 elements each. So, logically _MM_SHUFFLE should have taken 8 parameters. Can someone please explain how this works?

Answer 1

_mm256_shuffle_ps shuffles each of the two 128-bits lanes independently, as if _mm_shuffle_ps is called upon two XMM. If you want to shuffle all 8 32-bits elements, you need _mm256_permutevar8x32_ps