* `roll32 $aDst0:Dst0+1, $aSrc0:Src0+1, $aSrc1:Src1+1`

Perform a SIMD *roll* permutation on the 4 x 32-bit values across 2
source registers-pairs.