Unpack Cols in AVX2, C++ (gcc)

Language: Layout:

#include <stdio.h>
#include <immintrin.h>

void unpack_cols(__m256i lo, __m256i hi, __m256i& colA, __m256i& colB) {
    // group cols within lanes: 
    // a0 a1 b0 b1 a2 a3 b2 b3
    // a4 a5 b4 b5 a6 a7 b6 b7
    auto lo_shuffled = _mm256_shuffle_epi32(lo, _MM_SHUFFLE(3, 1, 2, 0));
    auto hi_shuffled = _mm256_shuffle_epi32(hi, _MM_SHUFFLE(3, 1, 2, 0));
    
    // unpack lo + hi a 64 bit
    // a0 a1 a4 a5 a2 a3 a6 a7
    // b0 b1 b4 b5 b2 b3 b6 b7
    auto colA_shuffled = _mm256_unpacklo_epi64(lo_shuffled, hi_shuffled);
    auto colB_shuffled = _mm256_unpackhi_epi64(lo_shuffled, hi_shuffled);
    
    // swap crossing lanes: 
    // a0 a1 a2 a3 a4 a5 a6 a7
    // b0 b1 b2 b3 b4 b5 b6 b7
    colA = _mm256_permute4x64_epi64(colA_shuffled, _MM_SHUFFLE(3, 1, 2, 0));
    colB = _mm256_permute4x64_epi64(colB_shuffled, _MM_SHUFFLE(3, 1, 2, 0));
}

int main()
{
    // Performance for Haswell
    
    // load the data: 
    // a0 b0 a1 b1 a2 b2 a3 b3
    // a4 b4 a5 b5 a6 b6 a7 b7
    auto lo = _mm256_setr_epi32(0xa0, 0xb0, 0xa1, 0xb1, 0xa2, 0xb2, 0xa3, 0xb3);
    auto hi = _mm256_setr_epi32(0xa4, 0xb4, 0xa5, 0xb5, 0xa6, 0xb6, 0xa7, 0xb7);
  
    __m256i colA, colB;
    unpack_cols(lo, hi, colA, colB);
    
    // dump col A
    // a0 a1 a2 a3 a4 a5 a6 a7
    printf("%02x %02x %02x %02x %02x %02x %02x %02x\n",
          _mm256_extract_epi32(colA, 0),
          _mm256_extract_epi32(colA, 1),
          _mm256_extract_epi32(colA, 2),
          _mm256_extract_epi32(colA, 3),
          _mm256_extract_epi32(colA, 4),
          _mm256_extract_epi32(colA, 5),
          _mm256_extract_epi32(colA, 6),
          _mm256_extract_epi32(colA, 7)          
    );
    
    // dump col B
    // b0 b1 b2 b3 b4 b5 b6 b7
    printf("%02x %02x %02x %02x %02x %02x %02x %02x\n",
          _mm256_extract_epi32(colB, 0),
          _mm256_extract_epi32(colB, 1),
          _mm256_extract_epi32(colB, 2),
          _mm256_extract_epi32(colB, 3),
          _mm256_extract_epi32(colB, 4),
          _mm256_extract_epi32(colB, 5),
          _mm256_extract_epi32(colB, 6),
          _mm256_extract_epi32(colB, 7)          
    );
}

Show compiler warnings [ + ] Compiler args [ + ] Show input
Compilation time: 0.42 sec, absolute running time: 0.14 sec, cpu time: 0 sec, memory peak: 3 Mb, absolute service time: 0,56 sec