Eduard S.
Eduard S.
Some files are quite big; we could try to split them to have a more manageable code base. Candidate list: - [halo2_frontend/src/circuit.rs](https://github.com/privacy-scaling-explorations/halo2/pull/290/files/7b44b19579a97c9e2ed059098e4334a8afc64dab#diff-0e169fae906b996a10f9bd9a3ef3300f75683600792086d52fee37b409496527): https://github.com/privacy-scaling-explorations/halo2/pull/290#discussion_r1522790030
Recently @han0110 did an analysis of peak memory estimation for halo2 proving, with a script implementation that takes a `ConstraintSystem` and estimates the peak memory of the prover. I ported...
In the key generation we find the following TODO: https://github.com/privacy-scaling-explorations/halo2/blob/73408a140737d8336490452193b21f5a7a94e7de/halo2_proofs/src/plonk/keygen.rs#L360-L365 We need to figure out: 1. Is it worth doing this part more efficiently? In my opinion, if this part...
From here: https://github.com/privacy-scaling-explorations/halo2/blob/73408a140737d8336490452193b21f5a7a94e7de/halo2_proofs/src/plonk.rs#L123 In halo2, whenever an update changes the `VeriyingKey` type in a non forward-compatible manner we're supposed to bump the version. When reading a serialized `VerifyingKey`, older versions...
Currently the API for proving/verifying receives the instances as: https://github.com/privacy-scaling-explorations/halo2/blob/73408a140737d8336490452193b21f5a7a94e7de/halo2_proofs/src/plonk/prover.rs#L49 I think that the only way to have this type is to create a vector of slices, take it as...
The following types don't have any comments; it would be nice to have some notes on what each type is: https://github.com/privacy-scaling-explorations/halo2/blob/73408a140737d8336490452193b21f5a7a94e7de/halo2_proofs/src/plonk/permutation/prover.rs#L21-L42
With https://github.com/privacy-scaling-explorations/halo2/pull/220 we have the option to mark some columns as unblinded. Nevertheless, even if all columns are unblinded, we're still leaving unusable rows: https://github.com/privacy-scaling-explorations/halo2/blob/0c3e3b569519b86653a64f412bbce17e4e8acac4/halo2_proofs/src/plonk/prover.rs#L300 Discussion from https://github.com/privacy-scaling-explorations/halo2/pull/243#discussion_r1452121055
The halo2 original implementation forces the circuit developer to choose a `FloorPlanner` when implementing the `Circuit` trait. Should we change the API so that the user of the circuit can...
During the review of https://github.com/privacy-scaling-explorations/halo2/pull/243 @CPerezz found various design choices from the original halo2 that could be improved: - A gate has a name, but also each constraint can have...
Follow up from https://github.com/privacy-scaling-explorations/zkevm-circuits/issues/1798 Estimate the rows needed for a block that tries to max out the blake and sha256 precompiles in a block with N gas. Sha256 is not...