Ok, so this might seem stupid.. but I am just trying to get my head around it.

I have 32bit registers (longword size).

Unsigned values are values from 0 to 255.

Adding two unsigned 128 bit values...

So do I need to use 4 registers per value and add the lower parts first, then the next lowest until the highest part?

Sure, just use the "add with carry" instruction, and you can add as many as you like.