Torbjörn Granlund tg at gmplib.org
Mon Aug 26 14:36:26 UTC 2019

I wonder if not something like this,

       for (;;) {
         if (up[N-1] < vp[N-1])
	   swap up,vp
	 cy = mpn_sub_n (up, up, vp, N);
	 if (UNLIKELY (cy)) { swap up,vp; goto back }
	 if (UNLIKELY (up[0] == 0)) {
	   // handle any number zeros including that U = 0
	 count_trailing_zeros (cnt, up[0]);
	 mpn_rshift (up, up, N, cnt);

would fit between our gcd_NN and the hgcd2 code?

The calls to mpn_sub_n and mpn_rshift should probably really be inlined
for specific values of N, but generic code for variable N might also be

Please encrypt, key id 0xC8601622

More information about the gmp-devel mailing list