Intel AVX: 256-bits version of dot product for double precision floating point variables

I would use a 4*double multiplication, then a hadd (which unfortunately adds only 2*2 floats in the upper and lower half), extract the upper half (a shuffle should work equally, maybe faster) and add it to the lower half.

The result is in the low 64 bit of dotproduct.

__m256d xy = _mm256_mul_pd( x, y );
__m256d temp = _mm256_hadd_pd( xy, xy );
__m128d hi128 = _mm256_extractf128_pd( temp, 1 );
__m128d dotproduct = _mm_add_pd( (__m128d)temp, hi128 );

Edit:
After an idea of Norbert P. I extended this version to do 4 dot products at one time.

__m256d xy0 = _mm256_mul_pd( x[0], y[0] );
__m256d xy1 = _mm256_mul_pd( x[1], y[1] );
__m256d xy2 = _mm256_mul_pd( x[2], y[2] );
__m256d xy3 = _mm256_mul_pd( x[3], y[3] );

// low to high: xy00+xy01 xy10+xy11 xy02+xy03 xy12+xy13
__m256d temp01 = _mm256_hadd_pd( xy0, xy1 );   

// low to high: xy20+xy21 xy30+xy31 xy22+xy23 xy32+xy33
__m256d temp23 = _mm256_hadd_pd( xy2, xy3 );

// low to high: xy02+xy03 xy12+xy13 xy20+xy21 xy30+xy31
__m256d swapped = _mm256_permute2f128_pd( temp01, temp23, 0x21 );

// low to high: xy00+xy01 xy10+xy11 xy22+xy23 xy32+xy33
__m256d blended = _mm256_blend_pd(temp01, temp23, 0b1100);

__m256d dotproduct = _mm256_add_pd( swapped, blended );