_MM_TRANSPOSE4_EPI32
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
#ifndef _MM_TRANSPOSE4_EPI32
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[1], q[3], q[5], q[7]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);
_MM_TRANSPOSE4_EPI32(q[0], q[2], q[4], q[6]);