Feature/gmp

[tlepoint]

Add conversion functions from and back to gmp (mpz_t, mpz_class).

[carlosaguilarmelchor]

@tlepoint great job !!

Some questions/remarks:

• The comment gmp.hpp:50 should be line 40
• you should use mpz_init2( x, bits_in_moduli_product) in gmp.hpp at lines 175,
• gmp.hpp:187 multilying bits_in_moduli_product by 4 is way too much (specially for very large moduli), it would be better to add (sizeof(T) /* for the multiplication when lifting */ + sizeof(size_t) /* for the degree */)*8
• in the white paper we changed "shoup values" for the more standard "newton quotients" a %s/shoup/newton/gc %s/Shoup/Newton quotients/gc command would be nice
• have you verified if the reduction you do gmp.hpp:196 with newton quotients is faster than gmp's fdiv_r ?
• gmp.hpp:124,199 params.hpp:23,56,99 logkMaxNbModuli can be removed as we have a compile-time static log function (in meta.hpp). static_log2<params<T>::kMaxNbModuli>::value should replace them in gmp.hpp

[tlepoint]

Thanks a lot, I pushed some modifications.

• Comment moved
• Done!
• It was too much indeed, I changed it more carefully
• I need to change all the Shoup/shoup into Newton/newton. Will do that later. Where is the white paper?
• yes, it was faster
• I didn't know about the static_log2. It's way better thanks!

[carlosaguilarmelchor]

Everything is ok for me. Hmm by the white paper I meant the paper accepted to CT-RSA (which is not white ...).

@serge-sans-paille @aguinet merge ?

[tlepoint]

Thanks for your comments @serge-sans-paille!

I removed completely all the "length" arguments, and used std::array instead of pointers. It is much cleaner thank you.

[tlepoint]

Thanks @serge-sans-paille, I cleaned up the setters. Is it better now?

[carlosaguilarmelchor]

Ok all comments seem to be adressed. Anyone needs more time to check the merge ?

If no more comments (or delay demands) are done before the end of the week I'll merge the PR.

[serge-sans-paille]

@carlosaguilarmelchor @tlepoint I'm ok with current state, Just made a few non-mandatory suggestion ;-)

Thanks a lot @tlepoint for taking into account my remarks!

[tlepoint]

Merging! Thanks @carlosaguilarmelchor and @serge-sans-paille