If you have two equal-sized buffers and you need to swap the over in memory:

[X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9]
becomes
[Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9 X0 X1 X2 X3 X4 X5 X6 X7 X8 X9]
[download]

It's easy to use a loop that swaps one machine register-sized chunk at a time:

void swapBufferHalves( U8 *buf, U32 start, U32 length ) {
    register U64 *p1 = (U64*)&buf[ start ];
    register U64 *p2 = (U64*)&buf[ start + ( length >> 1 ) ];
    const U64 *pend = (U64*)(start + length);
    register U64 temp;

    assert( !( length & 1 ) );
    assert( ( (U64)buf & 7ull ) == 0 );

    while( p1 < pend ) {
         temp = *p1;
         *p1++ = *p2;
         *p2++ = temp;
    }
    return;
}
[download]

But how to deal with the situation where the second 'half' of the buffer is shorter than the first:

[X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7]
becomes
[Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7 X0 X1 X2 X3 X4 X5 X6 X7 X8 X9]
[download]

Intuition tells me that this ought to be possible using 2 temporary registers; but the logic won't come?