Arrabbiata: new poseidon gadget #3053
Conversation
It seems that with nextest, it uses debug instead of release, even if it is compiled with release
dannywillems
force-pushed
the
arrabbiata/implement-new-poseidon-gadget
branch
2 times, most recently
from
1431888 to
65d7b28
Compare
dannywillems
force-pushed
the
arrabbiata/implement-new-poseidon-gadget
branch
from
65d7b28 to
46f2c8b
Compare
| /// setup, with [crate::NUMBER_OF_COLUMNS] columns, we can compute 5 full | ||
| /// The following gadgets implement the Poseidon hash instance described in | ||
| /// the top-level documentation. In the current setup, with | ||
| /// [crate::NUMBER_OF_COLUMNS] columns, we can compute 5 full | ||
| /// rounds per row. | ||
| Poseidon, |
There was a problem hiding this comment.
Note: the old Poseidon gadget will be removed later.
By introducing incrementally the gadget PoseidonPermutation, I hope it will ease the reviewer's work.
| /// using "public inputs". | ||
| /// | ||
| /// We split the Poseidon gadget in 13 sub-gadgets, one for each set of 5 | ||
| /// permutations and one for the absorbtion. The parameteris the starting |
There was a problem hiding this comment.
in my mind the term permutation is usually used for the full poseidon, not a round
There was a problem hiding this comment.
Yes, good point!
Changing it in a quick follow-up (to avoid breaking the PR train you already approved).
| /// using "public inputs". | ||
| /// | ||
| /// We split the Poseidon gadget in 13 sub-gadgets, one for each set of 5 | ||
| /// permutations and one for the absorbtion. The parameteris the starting |
There was a problem hiding this comment.
| /// permutations and one for the absorbtion. The parameteris the starting | |
| /// permutations and one for the absorbtion. The parameter is the starting |
| Gadget::EllipticCurveScaling => 2, | ||
| Gadget::Poseidon => 3, | ||
| Gadget::PoseidonSpongeAbsorb => 4, | ||
| Gadget::PoseidonPermutation(starting_round) => 5 + starting_round / 5, |
There was a problem hiding this comment.
I don't like that we could have two different gadget with the same conversion to usize
I know it's not supposed to happens as Gadget::PoseidonPermutation(6) should not be created but it has the same int representation as Gadget::PoseidonPermutation(10)
In general, as previously commented I'd be happy if we can avoid a conversion to usize
There was a problem hiding this comment.
Good point!
What about enforcing starting_round being a multiple of five and smaller than the number of full rounds?
I'll think about the conversion to usize. It is mostly to get an index and a count. There might be a more elegant solution.
Following #3050