Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat: Implementing IPA optimisation #4363

Merged
merged 4 commits into from
Feb 19, 2024
Merged

feat: Implementing IPA optimisation #4363

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c823606
IPA update
Rumata888 Jan 31, 2024
fc11cba
Revert accidental changes
Rumata888 Jan 31, 2024
dc6cd04
Update comments to reflect scheme change
Rumata888 Feb 16, 2024
329ff0e
Merge branch 'master' into is/documenting_ipa
Rumata888 Feb 19, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 19 additions & 22 deletions barretenberg/cpp/src/barretenberg/commitment_schemes/ipa/ipa.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,7 @@
* @brief IPA (inner-product argument) commitment scheme class. Conforms to the specification
* https://hackmd.io/q-A8y6aITWyWJrvsGGMWNA?view.
*
*
*/
namespace bb {
template <typename Curve> class IPA {
Expand Down Expand Up @@ -141,32 +142,30 @@ template <typename Curve> class IPA {
const Fr round_challenge = transcript->get_challenge<Fr>("IPA:round_challenge_" + index);
const Fr round_challenge_inv = round_challenge.invert();

auto G_lo = GroupElement::batch_mul_with_endomorphism(
std::span{ G_vec_local.begin(), G_vec_local.begin() + static_cast<long>(round_size) },
round_challenge_inv);
auto G_hi = GroupElement::batch_mul_with_endomorphism(
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We still consider this G_hi even though it's multiplying by the inverse now?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

G_hi is referring to multiplication of the higher indices of the vector, not to the power the challenge is used in

std::span{ G_vec_local.begin() + static_cast<long>(round_size),
G_vec_local.begin() + static_cast<long>(round_size * 2) },
round_challenge);
round_challenge_inv);

// Update the vectors a_vec, b_vec and G_vec.
// a_vec_next = a_vec_lo * round_challenge + a_vec_hi * round_challenge_inv
// b_vec_next = b_vec_lo * round_challenge_inv + b_vec_hi * round_challenge
// G_vec_next = G_vec_lo * round_challenge_inv + G_vec_hi * round_challenge
// a_vec_next = a_vec_lo + a_vec_hi * round_challenge
// b_vec_next = b_vec_lo + b_vec_hi * round_challenge_inv
// G_vec_next = G_vec_lo + G_vec_hi * round_challenge_inv
run_loop_in_parallel_if_effective(
round_size,
[&a_vec, &b_vec, round_challenge, round_challenge_inv, round_size](size_t start, size_t end) {
for (size_t j = start; j < end; j++) {
a_vec[j] *= round_challenge;
a_vec[j] += round_challenge_inv * a_vec[round_size + j];
b_vec[j] *= round_challenge_inv;
b_vec[j] += round_challenge * b_vec[round_size + j];
a_vec[j] += round_challenge * a_vec[round_size + j];
b_vec[j] += round_challenge_inv * b_vec[round_size + j];
}
},
/*finite_field_additions_per_iteration=*/4,
/*finite_field_multiplications_per_iteration=*/8,
/*finite_field_inversions_per_iteration=*/1);
GroupElement::batch_affine_add(G_lo, G_hi, G_vec_local);
GroupElement::batch_affine_add(
std::span{ G_vec_local.begin(), G_vec_local.begin() + static_cast<long>(round_size) },
G_hi,
G_vec_local);
}

transcript->send_to_verifier("IPA:a_0", a_vec[0]);
Expand Down Expand Up @@ -196,7 +195,7 @@ template <typename Curve> class IPA {
// Compute C_prime
GroupElement C_prime = opening_claim.commitment + (aux_generator * opening_claim.opening_pair.evaluation);

// Compute C_zero = C_prime + ∑_{j ∈ [k]} u_j^2L_j + ∑_{j ∈ [k]} u_j^{-2}R_j
// Compute C_zero = C_prime + ∑_{j ∈ [k]} u_j^{-1}L_j + ∑_{j ∈ [k]} u_jR_j
auto pippenger_size = 2 * log_poly_degree;
std::vector<Fr> round_challenges(log_poly_degree);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

seems like we can just get rid of this entirely?

Copy link
Contributor

@lucasxia01 lucasxia01 Feb 19, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

never mind, we need them for multiplying with element_R

std::vector<Fr> round_challenges_inv(log_poly_degree);
Expand All @@ -211,8 +210,8 @@ template <typename Curve> class IPA {

msm_elements[2 * i] = element_L;
msm_elements[2 * i + 1] = element_R;
msm_scalars[2 * i] = round_challenges[i].sqr();
msm_scalars[2 * i + 1] = round_challenges_inv[i].sqr();
msm_scalars[2 * i] = round_challenges_inv[i];
msm_scalars[2 * i + 1] = round_challenges[i];
}

GroupElement LR_sums = bb::scalar_multiplication::pippenger_without_endomorphism_basis_points<Curve>(
Expand All @@ -222,31 +221,29 @@ template <typename Curve> class IPA {
/**
* Compute b_zero where b_zero can be computed using the polynomial:
*
* g(X) = ∏_{i ∈ [k]} (u_{k-i}^{-1} + u_{k-i}.X^{2^{i-1}}).
* g(X) = ∏_{i ∈ [k]} (1 + u_{k-i}^{-1}.X^{2^{i-1}}).
*
* b_zero = g(evaluation) = ∏_{i ∈ [k]} (u_{k-i}^{-1} + u_{k-i}. (evaluation)^{2^{i-1}})
* b_zero = g(evaluation) = ∏_{i ∈ [k]} (1 + u_{k-i}^{-1}. (evaluation)^{2^{i-1}})
*/
Fr b_zero = Fr::one();
for (size_t i = 0; i < log_poly_degree; i++) {
auto exponent = static_cast<uint64_t>(Fr(2).pow(i));
b_zero *= round_challenges_inv[log_poly_degree - 1 - i] +
(round_challenges[log_poly_degree - 1 - i] * opening_claim.opening_pair.challenge.pow(exponent));
b_zero *= Fr::one() + (round_challenges_inv[log_poly_degree - 1 - i] *
opening_claim.opening_pair.challenge.pow(exponent));
}

// Compute G_zero
// First construct s_vec
std::vector<Fr> s_vec(poly_degree);
run_loop_in_parallel_if_effective(
poly_degree,
[&s_vec, &round_challenges, &round_challenges_inv, log_poly_degree](size_t start, size_t end) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nice catch

[&s_vec, &round_challenges_inv, log_poly_degree](size_t start, size_t end) {
for (size_t i = start; i < end; i++) {
Fr s_vec_scalar = Fr::one();
for (size_t j = (log_poly_degree - 1); j != size_t(-1); j--) {
auto bit = (i >> j) & 1;
bool b = static_cast<bool>(bit);
if (b) {
s_vec_scalar *= round_challenges[log_poly_degree - 1 - j];
} else {
s_vec_scalar *= round_challenges_inv[log_poly_degree - 1 - j];
}
}
Expand Down
Loading