* circom: add external_inputs
* adapt new external_inputs interface to the FoldingScheme trait and Nova impl
* adapt examples to new FCircuit external_inputs interface
* add state_len & external_inputs_len params to CircomFCircuit
* add examples/circom_full_flow.rs
* merge the params initializer functions, clippy
* circom: move r1cs reading to FCircuit::new instead of each step
* CI/examples: add circom so it can run the circom_full_flow example
* Add a dedicated variant of `mat_vec_mul_sparse` for `NonNativeFieldVar`
* Switch to a customized in-circuit nonnative implementation for efficiency
* Comments and tests for `NonNativeUintVar`
* Make `CycleFoldCircuit` a bit smaller
* Faster trusted setup and proof generation by avoiding some nested LCs
* Check the remaining limbs in a more safe way
* Format
* Disable the non-native checks in tests again
* Clarify the group operation in `enforce_equal_unaligned`
* Explain the rationale behind non-native mat-vec multiplication
* Explain the difference with some other impls of `enforce_equal_unaligned`
* Format
* Reduce the number of constraints in `AugmentedFCircuit`
For the test `folding::nova::tests::test_ivc`
Before: 138240
After: 86756 (1.6x improvement)
Two notable optimization techniques:
1. Instead of allocating two witness variables `a, b` and enforce their equality by calling `a.conditional_enforce_equal(&b, &cond)`, we can avoid the allocation of `b` and directly set `b = a`. The former might be costly due to the checks in allocation and `conditional_enforce_equal`. See `nova/circuits.rs` for details.
2. Before this commit, `NonNativeFieldVar::to_constraint_field` was majorly called for generating the inputs (preimage) to hash functions. However, it turns out that the underlying conversion strategy (optimized for weight) is not optimal for reducing the length of hash preimage. We can go further by maximizing the number of bits per limb, thereby minimizing the preimage length. See `circuits/nonnative.rs` for details.
* Format
* Fix clippy warnings
* Move the comments to the right position
* Cleanup unnecessary code
* Treat (the digest of) `cf_U_i1` as an additional public input to `AugmentedFCircuit` for full soundness
* Fix the y-coordinate in the affine form of zero points
This in turn fixes the inconsistency between the digest of a constant affine point and that of a witness affine point in circuits.
* Set `cf_u_i1_x` to the correct value
* Fix the number of public inputs in dummy instance and witness
* Unify the logic behind `CycleFoldCommittedInstanceVar::hash` and `CycleFoldChallengeGadget::get_challenge_gadget`
* Add `ToConstraintFieldGadget` bound to `GC2`
* Remove unnecessary code used for debugging
* Make clippy and rustfmt happy
* Move conversion methods for `NonNativeFieldVar` to `folding/circuits/nonnative.rs`
* Simplify the check of zero coordinates
* Gracefully handle the result of `nonnative_field_var_to_constraint_field`
* Make clippy happy again
* Compute Decider's CM challenges in Groth16 circuit, link G16 & KZG proofs in Onchain Decider, refactor CommitmentScheme trait
- Refactor commitment package
- Refactor `Commitment` trait and the kzg, ipa, pedersen impls
- Add methods to prove & verify given challenges (not computing them in-method)
- Add KZG challenges computation in decider_eth_circuit
- Add cmE & cmW KZG proving & verification in DeciderEth
- Link Decider's Groth16 proof & KZG proofs data
- Fix point to bytes arkworks inconsistency
- Patch ark_curves to use a cherry-picked version with bn254::constraints & grumpkin for v0.4.0 (once arkworks v0.5.0 is released this will no longer be needed)
* DeciderEthCircuit: Add check eval=p(c) for E & W
The check is temporary disabled due
https://github.com/privacy-scaling-explorations/folding-schemes/issues/80,
but the public inputs and logic are there, to be able to continue the
other parts development while issue #80 is solved.
* Change CycleFold approach:
Instead of having a single CycleFold circuit that checks the 2 forign
scalarmul of the main circuit instances, now there are 2 separated
CycleFold circuits each of them checking a single foreign scalarmul.
Increasing the number of constraints of the AugmentedFCircuit, but
reducing the number of constraints in the CycleFold circuit, which will
translate into reducing the number of constraints in the Decider
circuit.
* CycleFold circuits checks in AugmentedFCircuit:
- update NonNativeAffineVar to work with NonNativeFieldVar directly
instead of FpVar comming from NonNativeFieldVar.to_constraint_field()
- include in AugmentedFCircuit intermediate steps inbetween CycleFold
circuits, and update the internal checks of the CycleFold circuits
Pending to document the new CycleFold circuits approach and better
variable namings, rm unwraps, etc
* matrix_vec_mul_sparse gadget: skip value * v[col_i] mul when value==1
Saves a notable amount of constraints since there is a notable amount of
1 values in R1CS matrices.
* Reuse computed vector of U_i
Reuse computed vector of U_i, saving 4k constraints in AugmentedFCircuit.
* fixes post last rebase to main
* rm test_augmentedfcircuit since it is already tested in test_ivc (and is a slow computation)
* rm dbg!()
* small fixes after last main rebase
* Add IPA commitment native implementation
* Add IPA Gadget verifier
* polish Pedersen & IPA, add blind bool param to IPA
* Optimize IPA gadget constraints (and native):
- optimize <s,b> computation from linear to log time
- optimize s computation from k*2^k to k*(2^k)/2
* add small optimization: delegate u_i^-1 to prover and just check u_i*u_i^-1==1 in verifier circuit
* IPA polish and document
* Add 'BLIND' parameter to CommitmentProver trait (and to Pedersen and KZG impls). Fit IPA into CommitmentProver trait.
* rename 'BLIND' to 'H' (hiding) in commitment
* IPA: rm u_invs from Proof and compute them incircuit
* Update IPA's build_s & gadget to use Halo2 approach following @han0110 's suggestion.
This reduced further the amount of constraints needed.
- for k=4: -9k constraints (-7%)
- for k=8: -473k constr (-31%)
- for k=9: -1123k constr (-35%)
- for k=10: -2578k constr (-39%)
And now IPA verification (without amortizing) is very close to Pedersen
verification (in-circuits).
* rm dbg!(cs.num_constraints()) from multiple tests
* IPA::prove remove intermediate v_lo,v_hi vectors, add doc to build_s_gadget
* move powers_of into utils/mod.rs, update iters to cfg_iter
* Fix Nova multi-elements state
In the AugmentedFCircuit the default value for the state when no input
is provided was `vec![F::zero()]`, which defaults to length `1`. So when
having more than 1 element in the state, before even starting to fold,
the circuit was already already failing.
Additionally this commit adds an example for a circuit with a state of 5
elements.
* abstract 'nova_setup' helper to avoid code duplication in examples
* update example naming to 'MultiInputs'
* rename nova_setup -> test_nova_setup to make it more explicit
* change: Refactor structure into workspace
* chore: Add empty readme
* change: Transform repo into workspace
* add: Create folding-verifier-solidity crate
* add: Include askama.toml for `sol` extension escaper
* add: Jordi's old Groth16 verifier .sol template and adapt it
* tmp: create simple template struct to test
* Update FoldingSchemes trait, fit Nova+CycleFold
- update lib.rs's `FoldingScheme` trait interface
- fit Nova+CycleFold into the `FoldingScheme` trait
- refactor `src/nova/*`
* chore: add serialization assets for testing
Now we include an `assets` folder with a serialized proof & vk for tests
* Add `examples` dir, with Nova's `FoldingScheme` example
* polishing
* expose poseidon_test_config outside tests
* change: Refactor structure into workspace
* chore: Add empty readme
* change: Transform repo into workspace
* add: Create folding-verifier-solidity crate
* add: Include askama.toml for `sol` extension escaper
* add: Jordi's old Groth16 verifier .sol template and adapt it
* tmp: create simple template struct to test
* feat: templating kzg working
* chore: add emv and revm
* feat: start evm file
* chore: add ark-poly-commit
* chore: move `commitment` to `folding-schemes`
* chore: update `.gitignore` to ignore generated contracts
* chore: update template with bn254 lib on it (avoids import), update for loop to account for whitespaces
* refactor: update template with no lib
* feat: add evm deploy code, compile and create kzg verifier
* chore: update `Cargo.toml` to have `folding-schemes` available with verifiers
* feat: start kzg prove and verify with sol
* chore: compute crs from kzg prover
* feat: evm kzg verification passing
* tmp
* change: Swap order of G2 coordinates within the template
* Update way to serialize proof with correct order
* chore: update `Cargo.toml`
* chore: add revm
* chore: add `save_solidity`
* refactor: verifiers in dedicated mod
* refactor: have dedicated `utils` module
* chore: expose modules
* chore: update verifier for kzg
* chore: rename templates
* fix: look for binary using also name of contract
* refactor: generate groth16 proof for sha256 pre-image, generate groth16 template with verifying key
* chore: template renaming
* fix: switch circuit for circuit that simply adds
* feat: generates test data on the fly
* feat: update to latest groth16 verifier
* refactor: rename folder, update `.gitignore`
* chore: update `Cargo.toml`
* chore: update templates extension to indicate that they are templates
* chore: rename templates, both files and structs
* fix: template inheritance working
* feat: template spdx and pragma statements
* feat: decider verifier compiles, update test for kzg10 and groth16 templates
* feat: parameterize which size of the crs should be stored on the contract
* chore: add comment on how the groth16 and kzg10 proofs will be linked together
* chore: cargo clippy run
* chore: cargo clippy tests
* chore: cargo fmt
* refactor: remove unused lifetime parameter
* chore: end merge
* chore: move examples to `folding-schemes` workspace
* get latest main changes
* fix: temp fix clippy warnings, will remove lints once not used in tests only
* fix: cargo clippy lint added on `code_size`
* fix: update path to test circuit and add step for installing solc
* chore: remove `save_solidity` steps
* fix: the borrowed expression implements the required traits
* chore: update `Cargo.toml`
* chore: remove extra `[patch.crates-io]`
* fix: update to patch at the workspace level and add comment explaining this
* refactor: correct `staticcall` with valid input/output sizes and change return syntax for pairing
* refactor: expose modules and remove `dead_code` calls
* chore: update `README.md`, add additional comments on `kzg10` template and update `groth16` template comments
* chore: be clearer on attributions on `kzg10`
---------
Co-authored-by: CPerezz <c.perezbaro@gmail.com>
Co-authored-by: arnaucube <root@arnaucube.com>
* Add KZG commitment scheme adapted to vector commitment
Add KZG commitment scheme adapted to vector commitment
Also move the `src/pedersen.rs` into `src/commitment/pedersen.rs` where
it will coexist with `kzg.rs` and the trait defined in
`src/commitment/mod.rs`.
* Adapt Pedersen into the new CommitmentProver trait
* add CommitmentProver (Pedersen&KZG) homomorphic property test
* polishing
* Use divide_with_q_and_r, rename skip_first_zero_coeffs
Co-authored-by: han0110 <tinghan0110@gmail.com>
---------
Co-authored-by: han0110 <tinghan0110@gmail.com>
Changes:
- get rid of `extract_r1cs_and_z` and `extract_z`
- move `extract_r1cs` and `extract_w_x` from `frontend/arkworks` into
`r1cs.rs`
The reasoning: they are not methods needed for the Frontend interface,
but only needed internally for the folding scheme to extract values from
the AugmentedF circuit and similar.
- set the `FCircuit` as the trait for the `src/frontend`
- remove the `frontend/arkworks` since the `FCircuit` trait can be
directly implemented without a middle layer
- reorganize test circuits into `src/frontend/mod.rs`, updating them into
`CubicFCircuit`: the typical x^3+x+5=y circuit
`CustomFCircuit`: a circuit in which you can specify the number of
constraints that it will take
where both fulfill the `FCircuit` trait, and they are used for
different tests being folded.
* Implement Nova IVC's new & prove_step methods
Implement Nova IVC's new & prove_step methods (without CycleFold part yet)
* transcript.absorb_point err handling, and update C.xy() usage
* add transcript usage to IVC prove, add NovaTranscript trait extending Transcript trait, refactor NIFS.P to allow absorbing in transcript inbetween
* Implement Nova's IVC.V method (without CycleFold part yet)
* clippy lints
* move challenge r computation in-circuit
* reuse computed points with coordinates over CF (non-native) to save constraints in AugmentedFCircuit
(constraint count went down ~6k)
* rm 128 bit constant
* add params to Errors
* Updates from review suggestions. Additionally refactored nova/nifs fold, and rm transcript from nova/IVC.
- Updates from PR suggestions
- Additionally updated:
- in nova/nifs.rs: reuse folded_committed_instance for verify_folded_instance, computationally is the same, but reusing the same code so avoiding duplication and having an error on one of the two versions.
- in nova/ivc.rs: remove transcript from IVC (not needed, it uses the RO)
* impl AugmentedFCircuit non-base case
* add multiple iterations to AugmentedFCircuit test
* implement base case on AugmentedFCircuit and test
* Update cmE of E=0-vec to work as zero point
Update cmE of E=0-vec to work as zero point instead of as cm(0-vec)
* patch r1cs-std dep to a cherry-picked version with the zero-scalar-mult fix
* refactor FCircuit to make it more suitable inside the AugmentedFCircuit
* Implement Nova's NIFS.Verify circuits (with CycleFold)
- Add circuit for NIFS.Verify on the main curve to check the folded `u`
& `x`
- Add circuit for NIFS.Verify on the CycleFold's auxiliary curve to
check the folded `cm(E)` & `cm(W)`
- Add transcript.get_challenge_nbits
- Add tests for utils::vec.rs
* replace bls12-377 & bw6-761 by pallas & vesta curves (only affects tests)
We will use pallas & vesta curves (for tests only, the non-tests code
uses generics) for the logic that does not require pairings, and while
Grumpkin is not available
(https://github.com/privacy-scaling-explorations/folding-schemes/issues/12).
* update links to papers to markdown style
arnaucube
Ahmad Afuni
winderica
Pierre