Improving gmp

[coen dunnink]

Dear gmp,

I want to make an improvement to the gmp-libray. I did some libary-testing 
on integers with arbitrary precision(mpz_t) with the normal arithmetic 
functions. I noticed that the gmp-library was really the fastest library. I 
explicitly looked at the multiplication-operator. I want to make an 
improvement for multiplications  with to integers of the same length. Split 
them up and then multiply again. Like
Say 2 numbers x and y of each 100 bits. Then:
x * y = (2^50 * x1 + x2)  * (2^50 * y1 + y2)
     = (2^100 * x1 * y1) + (2^50 * x1 * y2) + (2^50 * x2 *y1) + ( x2 * y2)
This splitting and multiplying can only have effect on real big integers. 
But I wanted to know if there is already implemented such a thing. Else I 
like to improve and develope the gmp-library on this subject.

greetings
Coen Dunnink
Blekinge Tekniska Hogskola
Ronneby Sweden




_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com