feat!: Implement CommonReferenceString trait#139
Conversation
| witness: WitnessAssignments, | ||
| proving_key: &[u8], | ||
| ) -> Vec<u8> { | ||
| let circuit_size = self.get_circuit_size(constraint_system); |
There was a problem hiding this comment.
I do like this switch away from having to serialize a circuit to/from bberg for every composer method in order to fetch the circuit size and thus the data from the CRS. For methods such as create proof we have been forced to serialize the entire circuit to bberg to fetch the circuit size, and then once again to perform the main action of the method.
Is there anyway to persist the reference string in the Barretenberg struct once we call get_reference_string?
There was a problem hiding this comment.
I didn't like the idea of persisting it on the struct because then you have to do like Option<CRS> and check to see if you have it in each function. The idea behind get_reference_string is that it won't be called by nargo if Nargo already has a cached version that is large enough for the circuit, so we'd have to check/cache it in each function, which seems messy.
phated
force-pushed
the
phated/acvm-ref-string
branch
from
ce3cc59 to
81ad859
Compare
phated
force-pushed
the
phated/acvm-ref-string
branch
from
f4cc81b to
b28b147
Compare
phated
force-pushed
the
phated/acvm-ref-string
branch
from
b28b147 to
2cc62fd
Compare
phated
changed the title
phated
force-pushed
the
phated/acvm-ref-string
branch
from
0bdd53b to
84587ff
Compare
|
I'm see a lot of DNS failures on Linux. Do we know why or any way to avoid them? Is there a misconfiguration on the AWS bucket where the CRS lives? Or is it a problem with the Reqwest library (searching the internet for the error shows many similar errors from Reqwest...)? We wouldn't have seen these errors with the old implementation because we were caching in CI and then we were using Nix to fetch. |
phated
force-pushed
the
phated/acvm-ref-string
branch
from
67acf2c to
737d042
Compare
phated
force-pushed
the
phated/acvm-ref-string
branch
from
7d27ea0 to
5647a7f
Compare
* feat!: update to target acvm-84b5d18d * chore: update to use new black box solver interface * use patch syntax * update to latest changes * feat!: update to acvm with non-homogeneous bb calls (#169) * feat: update acvm * feat!: updated acvm to latest master * chore: update cargo toml * Update src/barretenberg_structures.rs Co-authored-by: Tom French <15848336+TomAFrench@users.noreply.github.com> --------- Co-authored-by: Tom French <15848336+TomAFrench@users.noreply.github.com> * fix bad rebase * feat!: Update backend to rely on WitnessMap (#152) Co-authored-by: Tom French <tom@tomfren.ch> * update acvm * feat!: Implement CommonReferenceString trait (#139) * chore!: Remove filesystem access and utilize async fetch with range requests Co-authored-by: Tom French <tom@tomfren.ch> * chore: Switch to tokio test macro for async function (#191) * chore: Remove `sled` & `tempfile` dev-dependencies (#190) * chore: Remove sled & tempfile dev-dependencies * chore: space out functions --------- Co-authored-by: Tom French <tom@tomfren.ch> * chore: bump to crates.io release * Update src/crs.rs * chore: Remove streaming CRS download & indicator (#194) * Update src/acvm_interop/common_reference_string.rs * code review --------- Co-authored-by: Blaine Bublitz <blaine.bublitz@gmail.com> Co-authored-by: Álvaro Rodríguez <sirasistant@gmail.com>
4 participants
This is a draft implementation of noir-lang/acvm#231 to see how the API worked out.
Closes #160
Closes #162