hyperledger-labs / fabric-token-sdk / 27709423171

82%

Build Type

push

github

Committed by AkramBitar

Commit Message

perf(ipa): native gnark-crypto dispatch for ComputeSVector and reduceVectors

Fixes #1618

The IPA prover and verifier inner loops (ComputeSVector, reduceVectors)
previously relied entirely on mathlib.Zr wrappers around big.Int, causing
O(n) heap allocations per butterfly round and significant GC pressure
during ZK proof generation and verification.

Changes:
- Add ipa_native.go with two generic functions:
    nativeComputeSVector[T, E]: dual-butterfly s-vector computation using
      in-place gnark-crypto field arithmetic, avoiding all big.Int
      allocations in the hot loop.
    nativeReduceVectors[T, E]: half-vector reduction using native field ops.
- Add curve-based dispatch in ComputeSVector and reduceVectors using the
  existing math.DispatchCurve helper:
    BLS12-381 -> bls12381fr.Element native path
    BN254     -> bn254fr.Element native path
    other     -> existing mathlib fallback (unchanged)
- Add ipa_native_test.go with:
    TestNativeComputeSVector_CrossCurve: inverse-property + definition
      consistency for BLS12-381 and BN254 across sizes 2..64.
    TestNativeIPAProofVerify: full prove→verify cycle on both curves.
    TestNativeReduceVectors_Consistency: IPA correctness after reduction.
    BenchmarkNativeComputeSVector / BenchmarkNativeIPAProver: alloc metrics.

All existing bulletproof tests continue to pass.

Signed-off-by: SurbhiAgarwal1 <agarwalsurbhi1807@gmail.com>

Coverage Stats

88 of 88 new or added lines in 2 files covered. (100.0%)

3 existing lines in 2 files now uncovered.

36322 of 43982 relevant lines covered (82.58%)

17.36 hits per line