Hi folks, I have written up some work I've been doing on middle products, please see http://cims.nyu.edu/~harvey/mulmid/ I expect this could yield improved constants in GMP for various styles of division, square root etc. Any comments, constructive or destructive, are welcome. david