I just have found out that a library that contains various wrapped ASM optimized procedures exists and that it is LGPL compatible and thus potentially interesting to (not only) allegro devs.
Maybe you know about it, maybe not, so sorry if you already know about it, but there is only one mention concerning that in the forum.
The library is called liboil (optimized inner loops) and you can read more about it here: http://liboil.freedesktop.org

Cool... I can stop re-inventing the wheel