Two-to-one aggregation circuits #216
Conversation
Bring back v0.3.0 into main + follow-up `trace_decoder` bump
github-actions
bot
added
the
crate: evm_arithmetization
| // Preprocess all circuits. | ||
| let all_circuits = AllRecursiveCircuits::<F, C, D>::new( | ||
| &all_stark, | ||
| &[8..17, 8..15, 8..18, 8..15, 8..10, 8..13, 8..20], |
There was a problem hiding this comment.
nit: let's at least use the start ranges of the empty_txn_list example so that it doesn't require a huge amount of RAM when trying it locally :)
| &[8..17, 8..15, 8..18, 8..15, 8..10, 8..13, 8..20], | |
| &[16..17, 9..15, 12..18, 14..15, 9..10, 12..13, 17..20], |
| pub two_to_one_aggregation: TwoToOneAggCircuitData<F, C, D>, | ||
| /// The block circuit, which verifies an aggregation root proof and an | ||
| /// optional previous block proof. | ||
| pub block: BlockCircuitData<F, C, D>, | ||
| /// The two-to-one block aggregation circuit, which verifies two unrelated | ||
| /// block proofs. | ||
| pub two_to_one_block: TwoToOneAggCircuitData<F, C, D>, |
There was a problem hiding this comment.
I'm a bit confused about the two_to_one_aggregation.
I see the use of aggregating two final block proofs together, but why aggregating intermediary layers that don't share anything?
There was a problem hiding this comment.
I think it could be useful in some cases. A "block proof" is just an "aggregation proof" along with a recursively verified previous block proof. But in some cases, the previous batch is already verified elsewhere (e.g. on L1), so the recursive proof is redundant.
I might be missing something though.
| pub two_to_one_aggregation: TwoToOneAggCircuitData<F, C, D>, | ||
| /// The block circuit, which verifies an aggregation root proof and an | ||
| /// optional previous block proof. | ||
| pub block: BlockCircuitData<F, C, D>, | ||
| /// The two-to-one block aggregation circuit, which verifies two unrelated | ||
| /// block proofs. | ||
| pub two_to_one_block: TwoToOneAggCircuitData<F, C, D>, |
There was a problem hiding this comment.
What exactly is the goal behind? This will only support one additional layer of aggregation, not an arbitrary one, which I would have thought was the desired goal?
There was a problem hiding this comment.
I think so too. I think this is just a temporary solution. @muursh ?
There was a problem hiding this comment.
Aye it is just a place holder for now. We will have Einar make some modifications to the current approach.
|
For the record, I am actually developing on this branch: https://github.com/0xPolygonZero/zk_evm/tree/two_to_one_aggregation_local_friendly |
|
Closing as agreed with @wborgeaud and @Nashtare. Superseded by #318. |
Add two circuits to
AllRecursiveCircuits:two_to_one_aggregationthat aggregates two unrelated aggregation proofs.two_to_one_blockthat aggregates two unrelated block proofs.