feat: Goblin translator opcode constraint and accumulator transfer relations (Goblin Translator part 5)

barretenberg/cpp/src/barretenberg/proof_system/relations/extra_relations.hpp

@@ -0,0 +1,188 @@
+#pragma once
+#include "barretenberg/numeric/uint256/uint256.hpp"
+#include "relation_parameters.hpp"
+#include "relation_types.hpp"
+
+namespace proof_system {
+
+template <typename FF_> class GoblinTranslatorOpcodeConstraintRelationImpl {
+  public:
+    using FF = FF_;
+
+    // 1 + polynomial degree of this relation
+    static constexpr size_t RELATION_LENGTH = 7; // degree(op(op - 1)(op - 2)(op - 3)(op - 4)(op - 8)) = 6
+    static constexpr std::array<size_t, 1> SUBRELATION_LENGTHS{
+        7 // opcode constraint relation
+    };
+
+    /**
+     * @brief Expression for enforcing the value of the Opcode to be {0,1,2,3,4,8}
+     * @details This relation enforces the opcode to be one of described values. Since we don't care about even
+     * values in the opcode wire and usually just set them to zero, we don't use a lagrange polynomial to specify
+     * the relation to be enforced just at odd indices, which brings the degree down by 1.
+     *
+     * @param evals transformed to `evals + C(in(X)...)*scaling_factor`
+     * @param in an std::array containing the fully extended Univariate edges.
+     * @param parameters contains beta, gamma, and public_input_delta, ....
+     * @param scaling_factor optional term to scale the evaluation before adding to evals.
+     */
+    template <typename ContainerOverSubrelations, typename AllEntitites>
+    void static accumulate(ContainerOverSubrelations& accumulators,
+                           const AllEntitites& in,
+                           const RelationParameters<FF>&,
+                           const FF& scaling_factor)
+    {
+
+        using Accumulator = std::tuple_element_t<0, ContainerOverSubrelations>;
+        using View = typename Accumulator::View;
+
+        auto op = View(in.op);
+        static const FF minus_one = FF(-1);
+        static const FF minus_two = FF(-2);
+        static const FF minus_three = FF(-3);
+        static const FF minus_four = FF(-4);
+        static const FF minus_eight = FF(-8);
+
+        // Contribution (1) (op(op-1)(op-2)(op-3)(op-4)(op-8))
+        auto tmp_1 = op * (op + minus_one);
+        tmp_1 *= (op + minus_two);
+        tmp_1 *= (op + minus_three);
+        tmp_1 *= (op + minus_four);
+        tmp_1 *= (op + minus_eight);
+        tmp_1 *= scaling_factor;
+        std::get<0>(accumulators) += tmp_1;
+    };
+};
+
+template <typename FF_> class GoblinTranslatorAccumulatorTransferRelationImpl {
+  public:
+    using FF = FF_;
+
+    // 1 + polynomial degree of this relation
+    static constexpr size_t RELATION_LENGTH = 3; // degree((SOME_LAGRANGE)(A-B)) = 2
+    static constexpr std::array<size_t, 12> SUBRELATION_LENGTHS{
+        3, // transfer accumulator limb 0 at even index subrelation
+        3, // transfer accumulator limb 1 at even index subrelation
+        3, // transfer accumulator limb 2 at even index subrelation
+        3, // transfer accumulator limb 3 at even index subrelation
+        3, // accumulator limb 0 is zero at the start of accumulation subrelation
+        3, // accumulator limb 1 is zero at the start of accumulation subrelation
+        3, // accumulator limb 2 is zero at the start of accumulation subrelation
+        3, // accumulator limb 3 is zero at the start of accumulation subrelation
+        3, // accumulator limb 0 is equal to given result at the end of accumulation subrelation
+        3, // accumulator limb 1 is equal to given result at the end of accumulation subrelation
+        3, // accumulator limb 2 is equal to given result at the end of accumulation subrelation
+        3  // accumulator limb 3 is equal to given result at the end of accumulation subrelation
+
+    };
+
+    /**
+     * @brief Relation enforcing non-arithmetic transitions of accumulator (value that is tracking the batched
+     * evaluation of polynomials in non-native field)
+     * @details This relation enforces three pieces of logic:
+     * 1) Accumulator starts as zero before we start accumulating stuff
+     * 2) Accumulator limbs stay the same if accumulation is not occurring (at even indices)
+     * 3) Accumulator limbs result in the values specified by relation parameters after accumulation
+     *
+     * @param evals transformed to `evals + C(in(X)...)*scaling_factor`
+     * @param in an std::array containing the fully extended Univariate edges.
+     * @param parameters contains beta, gamma, and public_input_delta, ....
+     * @param scaling_factor optional term to scale the evaluation before adding to evals.
+     */
+    template <typename ContainerOverSubrelations, typename AllEntities>
+    void static accumulate(ContainerOverSubrelations& accumulators,
+                           const AllEntities& in,
+                           const RelationParameters<FF>& relation_parameters,
+                           const FF& scaling_factor)
+    {
+        using Accumulator = std::tuple_element_t<0, ContainerOverSubrelations>;
+        using View = typename Accumulator::View;
+        // We use combination of lagrange polynomials at even indices in the minicircuit for copying the accumulator
+        auto lagrange_even = View(in.lagrange_even);
+
+        // Lagrange at index 1 is used to confirm the accumulator result
+        auto lagrange_second = View(in.lagrange_second);
+
+        // Lagrange at index (size of minicircuit - 2) is used to enforce that it starts with zero
+        auto lagrange_second_to_last_in_minicircuit = View(in.lagrange_second_to_last_in_minicircuit);
+
+        auto accumulators_binary_limbs_0 = View(in.accumulators_binary_limbs_0);
+        auto accumulators_binary_limbs_1 = View(in.accumulators_binary_limbs_1);
+        auto accumulators_binary_limbs_2 = View(in.accumulators_binary_limbs_2);
+        auto accumulators_binary_limbs_3 = View(in.accumulators_binary_limbs_3);
+        auto accumulators_binary_limbs_0_shift = View(in.accumulators_binary_limbs_0_shift);
+        auto accumulators_binary_limbs_1_shift = View(in.accumulators_binary_limbs_1_shift);
+        auto accumulators_binary_limbs_2_shift = View(in.accumulators_binary_limbs_2_shift);
+        auto accumulators_binary_limbs_3_shift = View(in.accumulators_binary_limbs_3_shift);
+
+        // Contribution (1) (1-4 ensure transfer of accumulator limbs at even indices of the minicircuit)
+        auto tmp_1 = accumulators_binary_limbs_0 - accumulators_binary_limbs_0_shift;
+        tmp_1 *= lagrange_even;
+        tmp_1 *= scaling_factor;
+        std::get<0>(accumulators) += tmp_1;
+
+        // Contribution (2)
+        auto tmp_2 = accumulators_binary_limbs_1 - accumulators_binary_limbs_1_shift;
+        tmp_2 *= lagrange_even;
+        tmp_2 *= scaling_factor;
+        std::get<1>(accumulators) += tmp_2;
+        // Contribution (3)
+        auto tmp_3 = accumulators_binary_limbs_2 - accumulators_binary_limbs_2_shift;
+        tmp_3 *= lagrange_even;
+        tmp_3 *= scaling_factor;
+        std::get<2>(accumulators) += tmp_3;
+        // Contribution (4)
+        auto tmp_4 = accumulators_binary_limbs_3 - accumulators_binary_limbs_3_shift;
+        tmp_4 *= lagrange_even;
+        tmp_4 *= scaling_factor;
+        std::get<3>(accumulators) += tmp_4;
+
+        // Contribution (5) (5-9 ensure that accumulator starts with zeroed-out limbs)
+        auto tmp_5 = accumulators_binary_limbs_0 * lagrange_second_to_last_in_minicircuit;
+        tmp_5 *= scaling_factor;
+        std::get<4>(accumulators) += tmp_5;
+
+        // Contribution (6)
+        auto tmp_6 = accumulators_binary_limbs_1 * lagrange_second_to_last_in_minicircuit;
+        tmp_6 *= scaling_factor;
+        std::get<5>(accumulators) += tmp_6;
+
+        // Contribution (7)
+        auto tmp_7 = accumulators_binary_limbs_2 * lagrange_second_to_last_in_minicircuit;
+        tmp_7 *= scaling_factor;
+        std::get<6>(accumulators) += tmp_7;
+
+        // Contribution (8)
+        auto tmp_8 = accumulators_binary_limbs_3 * lagrange_second_to_last_in_minicircuit;
+        tmp_8 *= scaling_factor;
+        std::get<7>(accumulators) += tmp_8;
+
+        // Contribution (9) (9-12 ensure the output is as stated, we basically use this to get the result out of the
+        // proof)
+        auto tmp_9 = (accumulators_binary_limbs_0 - relation_parameters.accumulated_result[0]) * lagrange_second;
+        tmp_9 *= scaling_factor;
+        std::get<8>(accumulators) += tmp_9;
+
+        // Contribution (10)
+        auto tmp_10 = (accumulators_binary_limbs_1 - relation_parameters.accumulated_result[1]) * lagrange_second;
+        tmp_10 *= scaling_factor;
+        std::get<9>(accumulators) += tmp_10;
+
+        // Contribution (11)
+        auto tmp_11 = (accumulators_binary_limbs_2 - relation_parameters.accumulated_result[2]) * lagrange_second;
+        tmp_11 *= scaling_factor;
+        std::get<10>(accumulators) += tmp_11;
+
+        // Contribution (12)
+        auto tmp_12 = (accumulators_binary_limbs_3 - relation_parameters.accumulated_result[3]) * lagrange_second;
+        tmp_12 *= scaling_factor;
+        std::get<11>(accumulators) += tmp_12;
+    };
+};
+template <typename FF>
+using GoblinTranslatorOpcodeConstraintRelation = Relation<GoblinTranslatorOpcodeConstraintRelationImpl<FF>>;
+
+template <typename FF>
+using GoblinTranslatorAccumulatorTransferRelation = Relation<GoblinTranslatorAccumulatorTransferRelationImpl<FF>>;
+
+} // namespace proof_system

barretenberg/cpp/src/barretenberg/proof_system/relations/goblin_translator_relation_consistency.test.cpp

@@ -13,9 +13,11 @@
  */
 #include "barretenberg/ecc/curves/bn254/fr.hpp"
 #include "barretenberg/proof_system/relations/decomposition_relation.hpp"
+#include "barretenberg/proof_system/relations/extra_relations.hpp"
 #include "barretenberg/proof_system/relations/gen_perm_sort_relation.hpp"
 #include "barretenberg/proof_system/relations/permutation_relation.hpp"
 #include "decomposition_relation.hpp"
+#include "extra_relations.hpp"
 #include <gtest/gtest.h>
 
 using namespace proof_system;
@@ -905,4 +907,77 @@ TEST_F(GoblinTranslatorRelationConsistency, DecompositionRelation)
     run_test(/*random_inputs=*/true);
 };
 
+TEST_F(GoblinTranslatorRelationConsistency, OpcodeConstraintRelation)
+{
+    const auto run_test = [](bool random_inputs) {
+        using Relation = GoblinTranslatorOpcodeConstraintRelation<FF>;
+        using RelationValues = typename Relation::ArrayOfValuesOverSubrelations;
+
+        const InputElements input_elements = random_inputs ? InputElements::get_random() : InputElements::get_special();
+        const auto& op = input_elements.op;
+
+        RelationValues expected_values;
+
+        const auto parameters = RelationParameters<FF>::get_random();
+
+        // (Contribution 1)
+        auto contribution_1 = op * (op - FF(1)) * (op - FF(2)) * (op - FF(3)) * (op - FF(4)) * (op - FF(8));
+        expected_values[0] = contribution_1;
+
+        validate_relation_execution<Relation>(expected_values, input_elements, parameters);
+    };
+    run_test(/*random_inputs=*/false);
+    run_test(/*random_inputs=*/true);
+};
+
+TEST_F(GoblinTranslatorRelationConsistency, AccumulatorTransferRelation)
+{
+    const auto run_test = [](bool random_inputs) {
+        using Relation = GoblinTranslatorAccumulatorTransferRelation<FF>;
+        using RelationValues = typename Relation::ArrayOfValuesOverSubrelations;
+
+        const InputElements input_elements = random_inputs ? InputElements::get_random() : InputElements::get_special();
+
+        const auto& lagrange_even = input_elements.lagrange_even;
+        const auto& lagrange_second = input_elements.lagrange_second;
+        const auto& lagrange_second_to_last_in_minicircuit = input_elements.lagrange_second_to_last_in_minicircuit;
+        const auto& accumulators_binary_limbs_0 = input_elements.accumulators_binary_limbs_0;
+        const auto& accumulators_binary_limbs_0_shift = input_elements.accumulators_binary_limbs_0_shift;
+        const auto& accumulators_binary_limbs_1 = input_elements.accumulators_binary_limbs_1;
+        const auto& accumulators_binary_limbs_1_shift = input_elements.accumulators_binary_limbs_1_shift;
+        const auto& accumulators_binary_limbs_2 = input_elements.accumulators_binary_limbs_2;
+        const auto& accumulators_binary_limbs_2_shift = input_elements.accumulators_binary_limbs_2_shift;
+        const auto& accumulators_binary_limbs_3 = input_elements.accumulators_binary_limbs_3;
+        const auto& accumulators_binary_limbs_3_shift = input_elements.accumulators_binary_limbs_3_shift;
+
+        RelationValues expected_values;
+
+        const auto parameters = RelationParameters<FF>::get_random();
+
+        const auto [accumulated_result_0, accumulated_result_1, accumulated_result_2, accumulated_result_3] =
+            parameters.accumulated_result;
+
+        // Check transfer of accumulator at even indices
+        expected_values[0] = lagrange_even * (accumulators_binary_limbs_0 - accumulators_binary_limbs_0_shift);
+        expected_values[1] = lagrange_even * (accumulators_binary_limbs_1 - accumulators_binary_limbs_1_shift);
+        expected_values[2] = lagrange_even * (accumulators_binary_limbs_2 - accumulators_binary_limbs_2_shift);
+        expected_values[3] = lagrange_even * (accumulators_binary_limbs_3 - accumulators_binary_limbs_3_shift);
+
+        // Check the accumulator starts as zero
+        expected_values[4] = accumulators_binary_limbs_0 * lagrange_second_to_last_in_minicircuit;
+        expected_values[5] = accumulators_binary_limbs_1 * lagrange_second_to_last_in_minicircuit;
+        expected_values[6] = accumulators_binary_limbs_2 * lagrange_second_to_last_in_minicircuit;
+        expected_values[7] = accumulators_binary_limbs_3 * lagrange_second_to_last_in_minicircuit;
+
+        // Check the accumulator results in submitted value
+        expected_values[8] = (accumulators_binary_limbs_0 - accumulated_result_0) * lagrange_second;
+        expected_values[9] = (accumulators_binary_limbs_1 - accumulated_result_1) * lagrange_second;
+        expected_values[10] = (accumulators_binary_limbs_2 - accumulated_result_2) * lagrange_second;
+        expected_values[11] = (accumulators_binary_limbs_3 - accumulated_result_3) * lagrange_second;
+        validate_relation_execution<Relation>(expected_values, input_elements, parameters);
+    };
+    run_test(/*random_inputs=*/false);
+    run_test(/*random_inputs=*/true);
+};
+
 } // namespace proof_system::ultra_relation_consistency_tests

barretenberg/cpp/src/barretenberg/proof_system/relations/relation_parameters.hpp

@@ -1,5 +1,5 @@
 #pragma once
-
+#include <array>
 namespace proof_system {
 
 /**
@@ -8,6 +8,7 @@ namespace proof_system {
  * @tparam FF
  */
 template <typename FF> struct RelationParameters {
+    static const int NUM_BINARY_LIMBS_IN_GOBLIN_TRANSLATOR = 4;
     FF eta = FF(0);                        // Lookup
     FF beta = FF(0);                       // Permutation + Lookup
     FF gamma = FF(0);                      // Permutation + Lookup
@@ -18,6 +19,7 @@ template <typename FF> struct RelationParameters {
     // eccvm_set_permutation_delta is used in the set membership gadget in eccvm/ecc_set_relation.hpp
     // We can remove this by modifying the relation, but increases complexity
     FF eccvm_set_permutation_delta = 0;
+    std::array<FF, NUM_BINARY_LIMBS_IN_GOBLIN_TRANSLATOR> accumulated_result = { FF(0) }; // Goblin Translator
 
     static RelationParameters get_random()
     {
@@ -32,6 +34,10 @@ template <typename FF> struct RelationParameters {
         result.eccvm_set_permutation_delta = result.gamma * (result.gamma + result.beta_sqr) *
                                              (result.gamma + result.beta_sqr + result.beta_sqr) *
                                              (result.gamma + result.beta_sqr + result.beta_sqr + result.beta_sqr);
+        result.accumulated_result = {
+            FF::random_element(), FF::random_element(), FF::random_element(), FF::random_element()
+        };
+
         return result;
     }
 };