chore: translator circuit builder and flavor audit by suyash67 · Pull Request #19798 · AztecProtocol/aztec-packages

suyash67 · 2026-01-21T14:24:51Z

🧾 Audit Context

Audit of translator circuit builder and flavor, no logic changes.

🛠️ Changes Made

Minor simplifications in circuit builder and flavor:
- use a single lambda function to split limb into microlimbs
- fix code comments (or formatting changes)
- fixed Lagrange polynomial sizes
- Added more circuit builder tests

✅ Checklist

Audited all methods of the relevant module/class
Audited the interface of the module/class with other (relevant) components
Documented existing functionality and any changes made (as per Doxygen requirements)
Resolved and/or closed all issues/TODOs pertaining to the audited files
Confirmed and documented any security or other issues found (if applicable)
Verified that tests cover all critical paths (and added tests if necessary)
Updated audit tracking for the files audited (check the start of each file you audited)

📌 Notes for Reviewers

--

Replace four specialized lambda functions (split_standard_limb_into_micro_limbs, split_top_limb_into_micro_limbs, split_top_z_limb_into_micro_limbs, and split_relation_limb_into_micro_limbs) with a single generic split_limb_into_microlimbs function that handles all cases. Fix bug where partial microlimbs were not correctly storing both unshifted and shifted versions, which is required for proper range constraints and reconstruction. Remove unused constants: TOP_STANDARD_MICROLIMB_BITS, TOP_Z_MICROLIMB_BITS, TOP_QUOTIENT_MICROLIMB_BITS. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Improve readability of complex mathematical equations by: - Using clang-format off/on directives to preserve custom formatting - Aligning similar terms vertically for easier visual parsing - Grouping terms logically (accumulator, v powers, quotient, remainder) - Maintaining all original logic without any changes Equations formatted: - remainder computation (accumulator formula) - quotient_by_modulus computation - low_wide_relation_limb_part_1 (lowest limb of relation) - low_wide_relation_limb (combined low relation with SHIFT_1) - high_wide_relation_limb (high relation with multiple limb combinations) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Split high_wide_relation_limb into two parts following the same pattern as low_wide_relation_limb for better readability and consistency: - high_wide_relation_limb_part_1: combines low limb divided + limbs (0*2, 1*1, 2*0) - high_wide_relation_limb: adds limbs (0*3, 1*2, 2*1, 3*0) * SHIFT_1 Also improved formatting of low_wide_relation_limb_part_1 by placing the variable initialization on a separate line for clearer structure. This makes both relation limb computations follow a consistent pattern: part_1 = base terms full = part_1 + (cross terms) * SHIFT_1 Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Improve equation readability by placing each multiplication term on its own line for the relation limb computations: - low_wide_relation_limb: 12 terms, each on separate line - high_wide_relation_limb_part_1: 17 terms, each on separate line - high_wide_relation_limb: 22 terms, each on separate line This vertical layout makes it much easier to: - Visually scan and verify all terms - Identify patterns and symmetries - Spot missing or duplicate terms - Understand the mathematical structure at a glance Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Reorder all multiplication terms in relation limb equations to consistently place x_witnesses or v*_witnesses on the right side of each multiplication: Before: v_witnesses[i] * p_x_limbs[j] After: p_x_limbs[j] * v_witnesses[i] This standardization: - Improves visual consistency across all equations - Makes it easier to identify the pattern: (limb * witness) - Groups terms by their limb indices when scanning vertically - Maintains mathematical equivalence (multiplication is commutative) Applied to all four equations: - low_wide_relation_limb_part_1 - low_wide_relation_limb - high_wide_relation_limb_part_1 - high_wide_relation_limb Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Refactor microlimb splitting to use a single loop for all standard limbs (P_x, P_y, current_accumulator, quotient) instead of four separate loops. This improves code organization and readability while maintaining the same logic. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Add doc comments to split_limb_into_microlimbs and improve lay_limbs_in_row documentation for better code readability. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

…neck, should be fast.

suyash67 · 2026-01-21T14:48:13Z

+     * Always returns NUM_MICRO_LIMBS (6) elements, padding with zeros if needed.
     */
-    auto split_top_limb_into_micro_limbs = [](const Fr& limb, const size_t last_limb_bits) {
+    auto split_limb_into_microlimbs = [](const Fr& limb, const size_t num_bits) {


simplification: consolidated all separate lambda fns to split into microlimbs in one lambda function.

suyash67 · 2026-01-21T14:49:36Z

    // Check and insert x_hi and z_1 into wire 2
    BB_ASSERT_LTE(uint256_t(ultra_op.x_hi), MAX_HIGH_WIDE_LIMB_SIZE);
-    BB_ASSERT_LTE(uint256_t(ultra_op.z_1), MAX_LOW_WIDE_LIMB_SIZE);
+    BB_ASSERT_LTE(uint256_t(ultra_op.z_1), MAX_Z_LIMB_SIZE);


Minor fix: we want to assert $z_1, z_2 < 2^{128}$ (not $z_1, z_2 < 2^{68 \times 2}$

suyash67 · 2026-01-21T14:51:18Z

+}
+
+// Test with minimal operations (only required no-ops and random ops)
+TEST(TranslatorCircuitBuilder, MinimalOperations)


Added lot more test cases in circuit builder

suyash67 · 2026-01-21T14:51:58Z

                              ordered_extra_range_constraints_numerator, // column 0
                              lagrange_first,                            // column 1
                              lagrange_last,                             // column 2
-                              // TODO(https://github.com/AztecProtocol/barretenberg/issues/758): Check if one of these


removed old issue that's closed

suyash67 · 2026-01-21T14:52:36Z

            lagrange_first = Polynomial{ /*size*/ 1, /*virtual_size*/ circuit_size };
            lagrange_result_row = Polynomial{ /*size*/ 1, /*virtual_size*/ circuit_size, /*start_index*/ RESULT_ROW };
-            lagrange_even_in_minicircuit = Polynomial{ /*size*/ MINI_CIRCUIT_SIZE - RESULT_ROW,
+            lagrange_even_in_minicircuit = Polynomial{ /*size*/ MINI_CIRCUIT_SIZE - RESULT_ROW - NUM_MASKED_ROWS_END,


Fixes polynomial sizes (actual range of polynomials)

iakovenkos · 2026-01-21T17:51:39Z

-    };
+        // Fill in the full 14-bit microlimbs
+        for (size_t i = 0; i < num_full_micro_limbs; ++i) {
+            microlimbs[i] = uint256_t(limb).slice(i * MICRO_LIMB_BITS, (i + 1) * MICRO_LIMB_BITS);


I think we can re-use faster slicing method here -- AztecProtocol/barretenberg#1610

Though maybe it requires non-Montgomery form. I'll check

Seems as per this thread, the function get_bit_slice_raw works for specifically for the Pippenger use cases.

iakovenkos · 2026-01-22T08:51:09Z

    // length = 3.
    // The degree has to be further increased because the relation is multiplied by the Row Disabling Polynomial
+    // total degree = sumcheck relation degree + 1 (PowZeta) + 1 (Lagrange)
    static constexpr size_t BATCHED_RELATION_PARTIAL_LENGTH = MAX_PARTIAL_RELATION_LENGTH + 2;


Shouldn't it be +1 and not +2 as we are not multiplying by Row Disabling Polynomial in Translator's sumcheck, and the masking Lagrange contributions are accounted by subrelations' lengths?

Opened a new issue AztecProtocol/barretenberg#1612

iakovenkos

LGTM!

BEGIN_COMMIT_OVERRIDE feat: support JSON input files for bb verify command (#19800) fix: update bootstrap.sh to use new JSON field names chore: Update `index.js` so that `HAS_ZK` and `PUBLIC_INPUTS` variables must always be set in tests (#19884) chore: pippenger int audit (#19302) chore: deduplicate batch affine addition trick (#19788) chore: transcript+codec+poseidon2 fixes (#19419) chore!: explicitly constrain inputs and intermediate witnesses (#19826) fix: exclude nlohmann/json from WASM builds in json_output.hpp chore: translator circuit builder and flavor audit (#19798) Revert "fix: exclude nlohmann/json from WASM builds in json_output.hpp" Revert "feat: support JSON input files for bb verify command (#19800)" Revert "fix: update bootstrap.sh to use new JSON field names" END_COMMIT_OVERRIDE

### 🧾 Audit Context Audit of translator circuit builder and flavor, no logic changes. ### 🛠️ Changes Made - Minor simplifications in circuit builder and flavor: - use a single lambda function to split limb into microlimbs - fix code comments (or formatting changes) - fixed Lagrange polynomial sizes - Added more circuit builder tests ### ✅ Checklist - [x] Audited all methods of the relevant module/class - [x] Audited the interface of the module/class with other (relevant) components - [x] Documented existing functionality and any changes made (as per Doxygen requirements) - [x] Resolved and/or closed all issues/TODOs pertaining to the audited files - [x] Confirmed and documented any security or other issues found (if applicable) - [x] Verified that tests cover all critical paths (and added tests if necessary) - [ ] Updated audit tracking for the files audited (check the start of each file you audited) ### 📌 Notes for Reviewers -- --------- Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

suyash67 and others added 11 commits January 21, 2026 14:23

Add documentation to lambda functions for consistency

2f607ba

Add doc comments to split_limb_into_microlimbs and improve lay_limbs_in_row documentation for better code readability. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

final set of cosmetic changes.

ea6a87f

add new translator circuit tests.

019a45a

remove todo, checking 10 values to be zero isn't that big of a bottle…

4afc6d3

…neck, should be fast.

minor fixes in flavor.

ca2f63c

suyash67 commented Jan 21, 2026

View reviewed changes

Comment thread barretenberg/cpp/src/barretenberg/translator_vm/translator_circuit_builder.cpp

suyash67 commented Jan 21, 2026

View reviewed changes

suyash67 requested a review from iakovenkos January 21, 2026 14:55

iakovenkos reviewed Jan 21, 2026

View reviewed changes

Comment thread barretenberg/cpp/src/barretenberg/translator_vm/translator_circuit_builder.cpp Outdated

iakovenkos reviewed Jan 22, 2026

View reviewed changes

iakovenkos approved these changes Jan 22, 2026

View reviewed changes

addr review.

be04293

suyash67 merged commit 91fc66e into merge-train/barretenberg Jan 26, 2026
8 checks passed

suyash67 deleted the sb/translator-ckt-bldr branch January 26, 2026 12:43

AztecBot mentioned this pull request Jan 26, 2026

feat: merge-train/barretenberg #19887

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

chore: translator circuit builder and flavor audit#19798

chore: translator circuit builder and flavor audit#19798
suyash67 merged 12 commits into
merge-train/barretenbergfrom
sb/translator-ckt-bldr

suyash67 commented Jan 21, 2026 •

edited

Loading

Uh oh!

suyash67 Jan 21, 2026

Uh oh!

Uh oh!

suyash67 Jan 21, 2026 •

edited

Loading

Uh oh!

suyash67 Jan 21, 2026

Uh oh!

suyash67 Jan 21, 2026 •

edited

Loading

Uh oh!

suyash67 Jan 21, 2026

Uh oh!

iakovenkos Jan 21, 2026

Uh oh!

iakovenkos Jan 22, 2026

Uh oh!

suyash67 Jan 26, 2026

Uh oh!

Uh oh!

iakovenkos Jan 22, 2026

Uh oh!

iakovenkos Jan 22, 2026

Uh oh!

iakovenkos left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

suyash67 commented Jan 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🧾 Audit Context

🛠️ Changes Made

✅ Checklist

📌 Notes for Reviewers

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

suyash67 Jan 21, 2026 • 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!

Choose a reason for hiding this comment

Uh oh!

suyash67 Jan 21, 2026 • 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!

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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

iakovenkos left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

suyash67 commented Jan 21, 2026 •

edited

Loading

suyash67 Jan 21, 2026 •

edited

Loading

suyash67 Jan 21, 2026 •

edited

Loading