feat: fixed size decider without padding #16318

iakovenkos · 2025-08-12T10:29:37Z

Closes AztecProtocol/barretenberg#1158
Closes AztecProtocol/barretenberg#1310
Closes AztecProtocol/barretenberg#1283

netlify · 2025-08-13T10:03:23Z

✅ Deploy Preview for aztec-docs-temp-you-can-delete ready!

Name	Link
🔨 Latest commit	`ce573fb`
🔍 Latest deploy log	https://app.netlify.com/projects/aztec-docs-temp-you-can-delete/deploys/689c62e90843a200086c29d7
😎 Deploy Preview	https://deploy-preview-16318--aztec-docs-temp-you-can-delete.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify project configuration.

netlify · 2025-08-14T15:29:36Z

✅ Deploy Preview for barretenberg ready!

Name	Link
🔨 Latest commit	`6970a75`
🔍 Latest deploy log	https://app.netlify.com/projects/barretenberg/deploys/689e1a2679213f00084360cf
😎 Deploy Preview	https://deploy-preview-16318--barretenberg.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify project configuration.

iakovenkos · 2025-08-14T17:20:43Z

barretenberg/cpp/src/barretenberg/commitment_schemes/gemini/gemini.hpp

                                                            std::span<const Fr> multilinear_challenge,
-                                                            const Polynomial& A_0);
+                                                            const Polynomial& A_0,
+                                                            const bool& has_zk = false);


See the explanation below

iakovenkos · 2025-08-14T17:26:08Z

barretenberg/cpp/src/barretenberg/commitment_schemes/gemini/gemini_impl.hpp

 template <typename Curve>
 std::vector<typename GeminiProver_<Curve>::Polynomial> GeminiProver_<Curve>::compute_fold_polynomials(
-    const size_t log_n, std::span<const Fr> multilinear_challenge, const Polynomial& A_0)
+    const size_t log_n, std::span<const Fr> multilinear_challenge, const Polynomial& A_0, const bool& has_zk)


the main Gemini changes are here:
the prover computes the folds for the first log_n rounds as before, and either scales the final constant fold during the virtual rounds or places zero polys to the claims. The reason for the latter is that UltraZKVerifiers would still need the log_circuit_size to evaluate the RowDisablingPolynomial in the very end of sumcheck.
Note MegaZKRecrursive verifier can always run for 20 rounds, cause it's used in a single place.

fcarreiro · 2025-08-18T07:47:08Z

barretenberg/cpp/src/barretenberg/vm2/constraining/verifier.cpp


    // Execute Sumcheck Verifier
-    const size_t log_circuit_size = numeric::get_msb(circuit_size);
+    // const size_t log_circuit_size = numeric::get_msb(circuit_size);


Beware that we were in the process of making the AVM circuit size non const: #16332

Thanks @fcarreiro! Now the verifier does not need to know the size of the circuit, just an upper bound, which in your case will be given by MAX_AVM_TRACE_LOG_SIZE. The main change is that in "virtual" rounds the prover is sending valid univariates in Sumchecck and valid commitments/evals in the PCS. The overhead is
num virtual rounds * MAX_RELATION_PARTIAL_LENGTH * num of ops to evaluate the entire relation at a given point.
So the Prover's work stays O(max_poly_size).

IIUC if we used MAX_AVM_TRACE_LOG_SIZE then the proof size would still be the correct unpadded one?

Looked at Lucas' PR - I think the avm prover must still use the actual polynomial size when initializing Sumcheck and doing Gemini, cause it saves work and memory + doesn't affect the VK

…l/aztec-packages into si/genuine-sumcheck-padding

iakovenkos · 2025-08-18T18:38:34Z

barretenberg/cpp/src/barretenberg/sumcheck/sumcheck_round.hpp


        return Utils::scale_and_batch_elements(relation_evaluations, alphas);
    }
-


iakovenkos · 2025-08-18T18:40:02Z

barretenberg/cpp/src/barretenberg/vm2/constraining/recursion/goblin_avm_recursive_verifier.hpp

        std::shared_ptr<Goblin::Transcript> transcript = std::make_shared<Goblin::Transcript>();
        // Construct Mega proof \pi_M of the AVM recursive verifier circuit
        auto mega_proving_key = std::make_shared<DeciderProvingKey_<MegaFlavor>>(mega_builder);
+        // Detect when MEGA_AVM_LOG_N needs to be bumped.


Seems useful to keep track of this. Currently it's 21.

ledwards2225

Good stuff. Let's get this in

ledwards2225 · 2025-08-19T03:12:29Z

barretenberg/cpp/src/barretenberg/commitment_schemes/gemini/gemini_impl.hpp

+    const Fr u_last = multilinear_challenge[log_n - 1];
+    const Fr final_eval = last.at(0) + u_last * (last.at(1) - last.at(0));
+    Polynomial const_fold(1);
+    // Temporary fix: when we're running a zk proof, the verifier uses a `padding_indicator_array`. So the evals in


Do you see a path to improving this? Worth an issue?

Yeah, I think I know how to remove it

…i/genuine-sumcheck-padding

BEGIN_COMMIT_OVERRIDE chore: Clean up public input propagation in `acir_format` (#16411) feat: merge via append ecc ops in hiding kernel (#16338) chore: install clang20, upgrade wasi-sdk to 27 (#16414) chore: make clang20 default and clean up presets (#16306) chore: stdlib sha256 without packed_byte_array (#15912) fix: Revert "chore: stdlib sha256 without packed_byte_array (#15912)" fix: Correct active range for databus (#16381) fix: Revert "fix: Correct active range for databus (#16381)" feat: fixed size decider without padding (#16318) Revert "chore: make clang20 default and clean up presets (#16306)" Revert "chore: install clang20, upgrade wasi-sdk to 27 (#16414)" chore: document the ECCVM builders. (#15913) END_COMMIT_OVERRIDE

Closes AztecProtocol/barretenberg#1158 Closes AztecProtocol/barretenberg#1310 Closes AztecProtocol/barretenberg#1283 --------- Co-authored-by: ledwards2225 <l.edwards.d@gmail.com> Co-authored-by: Maddiaa <47148561+Maddiaa0@users.noreply.github.com>

iakovenkos added 2 commits August 11, 2025 22:19

dumb case works

664f0f8

nonzk sumcheck works

2b9bfff

iakovenkos self-assigned this Aug 12, 2025

gemini prover supports ebz

ce573fb

iakovenkos added 15 commits August 13, 2025 17:09

ebz works

2747d95

remove print statements

7efc713

getting there, logderivative needs to handled more carefully

b3d937e

ultra tests pass

f1edb1b

remove unused methods

20f6308

ultra recursive is failing

4afe94a

ultra rec verifier works but uses real log_n

1d4194e

no log_circuit_size in pub_inputs

cfed4b5

minimal but brittle zk fix

354f855

use 20 as virtual log_n for mega --> fixes civc

f1692a1

fix gemini tests

1126294

dyadic_size -> evaluation separator in grand prod

d2e6ef6

some icky const proof size fixes to make Ultra Rec Verifier work

7bded8c

some docs + update gate count

fd40a63

Merge branch 'next' into si/genuine-sumcheck-padding

17a3f4f

iakovenkos added 5 commits August 14, 2025 15:42

fix num expected gates

b61eb29

fix shplemini recursion test

6a88ce3

actually fix the test

b01a5d0

adjust constants in mock verifier inputs

66c7857

fix gcc(?) build

6970a75

iakovenkos changed the base branch from next to merge-train/barretenberg August 14, 2025 17:19

iakovenkos marked this pull request as ready for review August 14, 2025 17:19

iakovenkos commented Aug 14, 2025

View reviewed changes

ledwards2225 added 2 commits August 15, 2025 22:33

cleanup and update vks

75bc29b

use constant for perm argument value separator

273d7e5

ledwards2225 added the ci-full Run all master checks. label Aug 15, 2025

iakovenkos added 2 commits August 17, 2025 19:40

bump const_pg_log_n to match avm mega circuit size

264af54

fix GoblinRecursion

4f8c671

iakovenkos requested review from IlyasRidhuan, fcarreiro and jeanmon as code owners August 17, 2025 23:20

fcarreiro reviewed Aug 18, 2025

View reviewed changes

iakovenkos and others added 3 commits August 18, 2025 12:49

fix num gates in mega zk rec verifier + fix sol honk contracts

b5b6617

Merge branch 'merge-train/barretenberg' into si/genuine-sumcheck-padding

675bc7e

chore: exclude non-empty stderr lines in gas_report file (#16441)

3486714

Maddiaa0 requested a review from charlielye as a code owner August 18, 2025 15:13

iakovenkos added 2 commits August 18, 2025 18:36

clean-up, docs, added an issue + MEGA_AVM_LOG_N constant

35b944e

Merge branch 'si/genuine-sumcheck-padding' of github.com:AztecProtoco…

f1fd2cc

…l/aztec-packages into si/genuine-sumcheck-padding

iakovenkos commented Aug 18, 2025

View reviewed changes

iakovenkos requested a review from ledwards2225 August 18, 2025 18:41

Merge branch 'merge-train/barretenberg' into si/genuine-sumcheck-padding

3cc70f7

ledwards2225 approved these changes Aug 19, 2025

View reviewed changes

iakovenkos and others added 3 commits August 19, 2025 15:05

Merge remote-tracking branch 'origin/merge-train/barretenberg' into s…

3011387

…i/genuine-sumcheck-padding

update vks

6e61504

Merge branch 'merge-train/barretenberg' into si/genuine-sumcheck-padding

3771f23

iakovenkos merged commit 5ee5a1f into merge-train/barretenberg Aug 19, 2025
6 checks passed

iakovenkos deleted the si/genuine-sumcheck-padding branch August 19, 2025 22:20

AztecBot mentioned this pull request Aug 19, 2025

feat: merge-train/barretenberg #16454

Merged


		return Utils::scale_and_batch_elements(relation_evaluations, alphas);
		}

feat: fixed size decider without padding #16318

feat: fixed size decider without padding #16318

Uh oh!

Conversation

iakovenkos commented Aug 12, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

netlify bot commented Aug 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Deploy Preview for aztec-docs-temp-you-can-delete ready!

Uh oh!

netlify bot commented Aug 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Deploy Preview for barretenberg ready!

Uh oh!

iakovenkos Aug 14, 2025

Choose a reason for hiding this comment

Uh oh!

iakovenkos Aug 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fcarreiro Aug 18, 2025

Choose a reason for hiding this comment

Uh oh!

iakovenkos Aug 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fcarreiro Aug 18, 2025

Choose a reason for hiding this comment

Uh oh!

iakovenkos Aug 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

iakovenkos Aug 18, 2025

Choose a reason for hiding this comment

Uh oh!

iakovenkos Aug 18, 2025

Choose a reason for hiding this comment

Uh oh!

ledwards2225 left a comment

Choose a reason for hiding this comment

Uh oh!

ledwards2225 Aug 19, 2025

Choose a reason for hiding this comment

Uh oh!

iakovenkos Aug 19, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

iakovenkos commented Aug 12, 2025 •

edited

Loading

netlify bot commented Aug 13, 2025 •

edited

Loading

netlify bot commented Aug 14, 2025 •

edited

Loading

iakovenkos Aug 14, 2025 •

edited

Loading

iakovenkos Aug 18, 2025 •

edited

Loading

iakovenkos Aug 18, 2025 •

edited

Loading