using SSE to calculate the average of a set of numbers

The average of a group of numbers is one of the fundamental statistical descriptions of data.  It's also easy to calculate in Ansi.
The horizontalSum we have seen before.

__forceinline float horizontalSum_SSE2(const __m128 &mABCD)
{
    __m128 mCDCD = _mm_movehl_ps(mABCD, mABCD);
    __m128 mApCBpD = _mm_add_ps(mABCD, mCDCD);
    __m128 mBpD = _mm_shuffle_ps(mApCBpD, mApCBpD, 0x55);
    __m128 mApBpCpD = _mm_add_ps(mApCBpD, mBpD);
    return _mm_cvtss_f32(mApBpCpD);
}
float average_SSE2(const float * const pA, const UINT uiAOrig)
{
    __m128 mSummed = _mm_setzero_ps();
    UINT uiA = uiAOrig;
    if (uiA & 1)
    {
        uiA--;
        mSummed = _mm_load_ss(&pA[uiA]);
    }
    if (uiA & 2)
    {
        uiA -= 2;
        mSummed = _mm_loadh_pi(mSummed, (const __m64*)&pA[uiA]);
    }
    while (uiA > 0)
    {
        uiA -= 4;
        mSummed = _mm_add_ps(mSummed, _mm_loadu_ps(&pA[uiA]));
    }
    return horizontalSum_SSE2(mSummed)/ uiAOrig;
}

There's not much new here.  But I will mention the line

mSummed = _mm_loadh_pi(mSummed, (const __m64*)&pA[uiA]);

we load into the other half of the xmm register than was loaded in the first load.  It's also sometimes necessary to cast among types for loads and other operations.