> Avoid using memcpy() for copying 32bit chunks. This shrinks > the resulting code a little. I think I put the memcpy code in when I was porting to ARM to reduce the diffs since it is needed there... Warner