Closed autholykos closed 5 years ago
Benchmarks confirm the expected performance improvement on signature generation using dusk-network/bn256@lattices compared to dusk-network/bn256@master.
*Specs
system ZenBook S UX391UA
bus UX391UA
memory 64KiB BIOS
memory 16GiB System Memory LPDDR3
processor Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
bridge Xeon E3-1200 v6/7th Gen Core Processor Host Bridge/DRAM Registers
Master
BenchmarkSign-8 3000 391747 ns/op
BenchmarkAggregate10Signatures-8 300000 4370 ns/op
BenchmarkAggregate100Signatures-8 30000 47491 ns/op
BenchmarkAggregate1000Signatures-8 3000 492140 ns/op
BenchmarkVerifySingleSignature-8 500 3603258 ns/op
BenchmarkVerifyUnsafeSingleSignature-8 500 3128698 ns/op
BenchmarkVerifySingleCompressedSignature-8 500 3199448 ns/op
BenchmarkVerifySingleCompressedUnsafeSignature-8 500 3000716 ns/op
Lattices
BenchmarkSign-8 5000 254454 ns/op
BenchmarkAggregate10Signatures-8 300000 4337 ns/op
BenchmarkAggregate100Signatures-8 30000 47310 ns/op
BenchmarkAggregate1000Signatures-8 3000 496335 ns/op
BenchmarkVerifySingleSignature-8 500 3398603 ns/op
BenchmarkVerifyUnsafeSingleSignature-8 500 3073851 ns/op
BenchmarkVerifySingleCompressedSignature-8 500 3124125 ns/op
BenchmarkVerifySingleCompressedUnsafeSignature-8 500 3134376 ns/op
While the
lattices
branch of bn256 is not expected to provide a substantial speed-up in signature verification (since it uses pairing operation, unaffected by the lattice-decomposition), it is equally true that signature generation and aggregation might benefit substantially from a more efficient scalar multiplication.This task is to benchmarks against
master
andlattices
branches