chore: civc tidy 3 #16671

ledwards2225 · 2025-08-29T20:37:52Z

More ClientIvc cleanup:

use precomputed VK for hiding kernel rather than recomputing it at runtime
don't use structured trace for hiding kernel VK construction in mock circuit producer
remove use of non-static verify() and prove_and_verify() in favor of static ClientIvc::verify()
remove no longer needed propagation of tail public inputs in hiding kernel (it no longer has any!)
simplify ClientIvc member variables: remove honk_vk and replace ProverFoldOutput instance with a simple prover_accumulator

ledwards2225 · 2025-09-09T17:59:40Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.cpp

        hide_op_queue_accumulation_result(circuit);

-        // Propagate the public inputs of the tail kernel by converting them to public inputs of the hiding circuit.
-        auto num_public_inputs = static_cast<size_t>(honk_vk->num_public_inputs);


the tail now uses the databus - it has no public inputs to propagate

ah great! I think I removed it and had some tests failing in my previous clean up.

ledwards2225 · 2025-09-09T18:02:07Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.cpp

 ClientIVC::VerificationKey ClientIVC::get_vk() const
 {
-    return { honk_vk, std::make_shared<ECCVMVerificationKey>(), std::make_shared<TranslatorVerificationKey>() };
+    BB_ASSERT_EQ(verification_queue.size(), 1UL);


we now just extract the hiding kernel VK from the queue

ledwards2225 · 2025-09-09T18:03:13Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.hpp

  public:
    size_t num_circuits_accumulated = 0; // number of circuits accumulated so far

-    ProverFoldOutput fold_output; // prover accumulator and fold proof


we only track the accumulator since the proof is stored internal to the verification queue

ledwards2225 · 2025-09-09T18:03:41Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.test.cpp


        for (size_t j = 0; j < num_circuits; ++j) {
-            // Use default test settings for the mock hiding kernel since it's size must always be consistent
-            if (j == num_circuits - 1) {


this is now handled internal to the mock circuit produced

ledwards2225 · 2025-09-09T18:05:32Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.cpp

 {
    // Note: a structured trace is not used for the hiding kernel
    auto hiding_decider_pk = std::make_shared<DeciderZKProvingKey>(circuit, TraceSettings(), bn254_commitment_key);
-    honk_vk = std::make_shared<MegaZKVerificationKey>(hiding_decider_pk->get_precomputed());


prior to this PR we were recomputing the hiding VK on the fly rather than using the precomputed one. This is why we weren't getting errors even though the precomputed VK was incorrectly using a structured trace

ledwards2225 · 2025-09-09T18:05:55Z

barretenberg/cpp/src/barretenberg/client_ivc/test_bench_shared.hpp

- * @brief Verify an IVC proof
- *
- */
-bool verify_ivc(ClientIVC::Proof& proof, ClientIVC& ivc)


ledwards2225 · 2025-09-09T18:06:11Z

barretenberg/cpp/src/barretenberg/dsl/acir_format/pg_recursion_constraint.cpp

    ivc->goblin.merge_verification_queue.emplace_back(acir_format::create_mock_merge_proof());
    // If the type is PG_FINAL, we also need to populate the ivc instance with a mock decider proof
    if (type == ClientIVC::QUEUE_TYPE::PG_FINAL) {
-        // we have to create a mock honk vk


not anymore we dont

ledwards2225 · 2025-09-09T18:06:45Z

noir-projects/noir-protocol-circuits/bootstrap.sh


 # Circuits matching these patterns we have client-ivc keys computed, rather than ultra-honk.
 readarray -t ivc_patterns < <(jq -r '.[]' "../client_ivc_circuits.json")
-readarray -t ivc_tail_patterns < <(jq -r '.[]' "../client_ivc_tail_circuits.json")


just correcting some naming here from "tail" to "hiding"

kashbrti

LG! thanks for the clean up

kashbrti · 2025-09-10T09:34:32Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.cpp

        hide_op_queue_accumulation_result(circuit);

-        // Propagate the public inputs of the tail kernel by converting them to public inputs of the hiding circuit.
-        auto num_public_inputs = static_cast<size_t>(honk_vk->num_public_inputs);


ah great! I think I removed it and had some tests failing in my previous clean up.

kashbrti · 2025-09-10T09:36:17Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.cpp

    }
    auto verifier_instance = std::make_shared<DeciderVerificationKey_<Flavor>>(honk_vk);
-    FoldingProver folding_prover({ fold_output.accumulator, proving_key },
+    FoldingProver folding_prover({ prover_accumulator, proving_key },


quite funny. I think we were only using the accumulator of the fold_output but keeping the whole object. did we historically use another member of it?

the other member was the proof. Can't remember if we ever used it but wouldnt make much sense since its stored in the queue anyway

kashbrti · 2025-09-10T09:38:48Z

barretenberg/cpp/src/barretenberg/client_ivc/client_ivc.cpp

+
    // Hiding circuit is proven by a MegaZKProver
-    MegaZKProver prover(hiding_decider_pk, hiding_circuit_vk, transcript);
+    MegaZKProver prover(hiding_decider_pk, verification_key, transcript);


just to make sure I understand correctly. We do not keep the honk_vk member anymore (that use to keep the vk of the last proof that needs to be recursively verified in the hiding circuit) and we're just passing it as an input here?

yeah ever since the updates from you/leila that final VK (hiding VK) is a precomputed input just like for all of the other circuits - until now we just weren't using it and instead recomputed it on the fly

kashbrti · 2025-09-10T09:45:51Z

barretenberg/cpp/src/barretenberg/client_ivc/mock_circuit_producer.hpp

    std::pair<ClientCircuit, std::shared_ptr<VerificationKey>> create_next_circuit_and_vk(ClientIVC& ivc,
                                                                                          TestSettings settings = {})
    {
+        // If this is a mock hiding kernel, remove the settings and use a default (non-structured) trace


oops seems like I forgot about this when refactoring the tests. thanks!

BEGIN_COMMIT_OVERRIDE fix: Origin Tags edgecase (#16921) chore: cycle group cleanup #2 (#16876) chore: civc tidy 3 (#16671) refactor(bb): optimize batch_mul_with_endomorphism (#16905) feat: check op queue wires are zero past minicircuit in Translator (#16858) feat: Add CPU scaling benchmark script for remote execution (#16918) fix: Add free witness tag to field constructor (#16827) fix(bb): darwin build (#16957) END_COMMIT_OVERRIDE

More ClientIvc cleanup: - use precomputed VK for hiding kernel rather than recomputing it at runtime - don't use structured trace for hiding kernel VK construction in mock circuit producer - remove use of non-static `verify()` and `prove_and_verify()` in favor of static `ClientIvc::verify()` - remove no longer needed propagation of tail public inputs in hiding kernel (it no longer has any!) - simplify `ClientIvc` member variables: remove `honk_vk` and replace `ProverFoldOutput` instance with a simple `prover_accumulator`

ledwards2225 added 2 commits August 29, 2025 19:20

Progress towards not using global honk_vk

1e1aa97

remove honk_vk altogether

007db1e

ledwards2225 marked this pull request as ready for review August 29, 2025 20:42

ledwards2225 requested review from IlyasRidhuan, Maddiaa0, fcarreiro and jeanmon as code owners August 29, 2025 20:42

WiP replace fold_output with prover_accumulator

c016a25

ledwards2225 changed the base branch from next to merge-train/barretenberg August 29, 2025 20:48

ledwards2225 added 3 commits August 30, 2025 19:28

fix cast issue

ae13c4a

add worktodo

674d53a

remove redundant verify methods

07cac06

ledwards2225 added the ci-full Run all master checks. label Aug 30, 2025

ledwards2225 added 4 commits August 30, 2025 20:23

tidy

3e562a5

Merge branch 'merge-train/barretenberg' into lde/civc-tidy-3

65ba35d

fix assert

f9c8770

fix build

acc73bc

fcarreiro removed their request for review September 1, 2025 17:13

ledwards2225 added 5 commits September 2, 2025 15:40

Merge branch 'merge-train/barretenberg' into lde/civc-tidy-3

be7457c

restore randomly deleted components

063b712

tidy

cb3f22b

add option to not use structured trace when computing standalone vk

cb6e829

update cli to allow nonstructured hiding circuit VK

045a909

ledwards2225 requested a review from LeilaWang as a code owner September 2, 2025 22:11

ledwards2225 added 6 commits September 2, 2025 23:29

Merge branch 'merge-train/barretenberg' into lde/civc-tidy-3

8d9097b

nargo fmt

9dca508

again

6b792ed

Merge branch 'merge-train/barretenberg' into lde/civc-tidy-3

0308e4f

fmt?

824cf8c

unformat??

2989802

ledwards2225 added 2 commits September 3, 2025 03:59

again

c0f85e4

bump prover full timeout

17da7bb

ledwards2225 requested a review from charlielye as a code owner September 3, 2025 16:45

ledwards2225 added 9 commits September 3, 2025 21:52

test: bump timeout to 25

7e1d814

Merge branch 'merge-train/barretenberg' into lde/civc-tidy-3

c7bcd9f

wasm fix

5c1ed1a

remove no longer needed tail pub input propagation

8f089ef

touch more cleanup

0351a10

Merge branch 'merge-train/barretenberg' into lde/civc-tidy-3

be46a3f

Merge branch 'merge-train/barretenberg' into lde/civc-tidy-3

ec09536

update noir protocol boot

4ebd6f1

Merge branch 'merge-train/barretenberg' into lde/civc-tidy-3

0543364

ledwards2225 commented Sep 9, 2025

View reviewed changes

cleanup

578e8fe

ledwards2225 requested a review from kashbrti September 9, 2025 18:30

kashbrti approved these changes Sep 10, 2025

View reviewed changes

ledwards2225 merged commit 22602e5 into merge-train/barretenberg Sep 10, 2025
6 checks passed

ledwards2225 deleted the lde/civc-tidy-3 branch September 10, 2025 16:50

AztecBot mentioned this pull request Sep 10, 2025

feat: merge-train/barretenberg #16928

Merged

chore: civc tidy 3 #16671

chore: civc tidy 3 #16671

Uh oh!

Conversation

ledwards2225 commented Aug 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kashbrti left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

ledwards2225 commented Aug 29, 2025 •

edited

Loading