Module curves_perf

Available on docsrs only.

Curves performance

Below you can find benchmark results for all curves provided in generic-ec.

Capital letter stand for a point, and small letter stand for a scalar. E.g. A+B is point addition, a+b is scalar addition, and [k]P is multiplication of scalar at point.

	secp256k1	secp256r1	stark	ed25519
A+B	300ns	453ns	505ns	215ns
[k]P	52.5μs	140.8μs	153.5μs	36.6μs
SmallFactorCheck	8ns	5ns	6ns	36606ns
EncodeCompressedPoint	5.8μs	21.1μs	14.9μs	3.5μs
EncodeUncompressedPoint	5.8μs	21.1μs	14.9μs	3.5μs
DecodeCompressedPoint	5.7μs	7.0μs	2599.6μs	4.0μs
DecodeUncompressedPoint	129ns	225ns	336ns	3966ns
a+b	11ns	11ns	11ns	35ns
a*b	41ns	58ns	29ns	98ns
inv(a)	13.2μs	21.9μs	8.3μs	11.4μs
RandomScalar	23ns	23ns	1803ns	146ns
EncodeScalarBE	8ns	8ns	23ns	5ns
EncodeScalarLE	12ns	12ns	22ns	3ns
DecodeScalarBE	5ns	5ns	2853ns	114ns
DecodeScalarLE	4ns	4ns	2852ns	112ns
BeBytesModOrder/32	30ns	30ns	2811ns	79ns
LeBytesModOrder/32	29ns	29ns	2807ns	77ns
BeBytesModOrder/64	59ns	121ns	8423ns	148ns
LeBytesModOrder/64	46ns	120ns	8416ns	140ns
BeBytesModOrder/128	184ns	264ns	14077ns	566ns
LeBytesModOrder/128	166ns	261ns	14066ns	544ns
BeBytesModOrder/512	691ns	1126ns	48027ns	2143ns
LeBytesModOrder/512	626ns	1111ns	47862ns	2065ns
ReduceBe/32	13ns	13ns	2745ns	63ns
ReduceLe/32	12ns	12ns	2734ns	61ns
ReduceBe/64	42ns			130ns
ReduceLe/64	28ns			124ns

Reproducibility

Benchmarks were carried out on c5.2xlarge EC2 instance.

In order to reproduce the experiments, you need cargo-criterion to be installed. Then you can run benchmarks via:

cargo criterion --message-format json --bench curves > ./perf/curves/results.json

Then you can rebuild the table above:

cargo run --bin analyze -- curves < perf/curves/results.json > perf/curves/table.md