Nikos Mavrogiannopoulos nmav@gnutls.org writes:
I think there is benefit from Martin's patches irrespective of the LGPL argument. The system I'm using at work has an old gmp (not for license reasons, just an old distribution which is not up to me to upgrade)
Checked in now. I think I have all Martin's patches in now (let me know if I forgot something). It would be good if you or Martin could check if it works with older gmp releases now.
Other recent changes:
* Today, I wrote some documentation on the high-level ecc functions (basically, the highest level ecdsa functions and what's needed to use them). Comments appreciated.
* I added the ecc_point_mul and ecc_point_mul_g functions (but testcases missing).
* I added the salsa20r12 testvectors you sent me a while ago.
* I have arranged the ecc code to use gmp's mpn_cnd_{add,sub}_n functions when available (not yet in any release). Mixed benchmarking results on ARM; I use mpn_addmul_1 and mpn_submul_1 as fallbacks, and for small operands mpn_addmul_1 appears to be slightly faster than mpn_add_n on that machine, slowing some operations down a little.
Regards, /Niels