Pentium 4 SSE2 challenge

[Paul Zimmermann]

       Torbjo"rn,

Laurent Fousse forwarded me your Pentium 4 SSE2 challenge. Looks interesting,
but since my assembly knowledge is almost nothing, I won't participate.

However, we could propose the following challenge to hardware manufacturers:
implement a processor with an operation (hi,lo) <- a*b+c+d, where all operands
have one-limb. With such an operation, it would be theoretically possible to
reach 1.0 cycle/limb for mpn_addmul_1.

Paul