saymrwulf/risc0-curve25519-dalek-source
1
0
Fork 0
mirror of https://github.com/saymrwulf/risc0-curve25519-dalek-source.git synced 2026-09-06 20:20:39 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Update benchmarks in README.

Browse source
This commit is contained in:
Isis Lovecruft 2018-07-15 23:22:22 +00:00
parent b32e39c801
commit 5050049a6c
Failed to extract signature
1 changed files with 28 additions and 27 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

55
README.md
View file

@ -14,46 +14,47 @@ reason for feature-gating the benchmarks is that Rust's `test::Bencher` is
unstable, and thus only works on the nightly channel. (We'd like people to be
able to compile and test on the stable and beta channels too!)
On an Intel i5 Sandy Bridge running at 2.6 GHz, with TurboBoost enabled (and
also running in QubesOS with *lots* of other VMs executing), this code
achieves the following performance benchmarks:
On an Intel i9-7900X running at 3.30 GHz, without TurboBoost, this code achieves
the following performance benchmarks:
∃!isisⒶwintermute:(master *=)~/code/rust/ed25519-dalek ∴ cargo bench --features="nightly bench"
∃!isisⒶmistakenot:(master *=)~/code/rust/ed25519-dalek ∴ cargo bench
Compiling ed25519-dalek v0.7.0 (file:///home/isis/code/rust/ed25519-dalek)
Finished release [optimized] target(s) in 3.11s
Running target/release/deps/ed25519_dalek-ae92163eefd0cc80
Running target/release/deps/ed25519_benchmarks-721332beed423bce
running 9 tests
test ed25519::test::golden ... ignored
test ed25519::test::public_key_from_bytes ... ignored
test ed25519::test::sign_verify ... ignored
test ed25519::test::unmarshal_marshal ... ignored
test ed25519::bench::key_generation ... bench: 30,711 ns/iter (+/- 10,936)
test ed25519::bench::sign ... bench: 39,432 ns/iter (+/- 21,387)
test ed25519::bench::sign_expanded_key ... bench: 45,753 ns/iter (+/- 25,261)
test ed25519::bench::underlying_scalar_mult_basepoint ... bench: 25,455 ns/iter (+/- 10,587)
test ed25519::bench::verify ... bench: 91,408 ns/iter (+/- 31,193)
Ed25519 signing time: [15.617 us 15.630 us 15.647 us]
Ed25519 signature verification time: [45.930 us 45.968 us 46.011 us]
Ed25519 keypair generation time: [15.440 us 15.465 us 15.492 us]
test result: ok. 0 passed; 0 failed; 4 ignored; 5 measured; 0 filtered out
By enabling the avx2 backend (on machines with compatible microarchitectures),
the performance for signature verification is greatly improved:
∃!isisⒶmistakenot:(master *=)~/code/rust/ed25519-dalek ∴ export RUSTFLAGS="-C target_cpu=native"
∃!isisⒶmistakenot:(master *=)~/code/rust/ed25519-dalek ∴ cargo bench --no-default-features --features "std avx2_backend"
Compiling ed25519-dalek v0.7.0 (file:///home/isis/code/rust/ed25519-dalek)
Finished release [optimized] target(s) in 4.28s
Running target/release/deps/ed25519_benchmarks-e4866664de39c84d
Ed25519 signing time: [15.923 us 15.945 us 15.967 us]
Ed25519 signature verification time: [33.382 us 33.411 us 33.445 us]
Ed25519 keypair generation time: [15.246 us 15.260 us 15.275 us]
In comparison, the equivalent package in Golang performs as follows:
∃!isisⒶwintermute:(master *=)~/code/go/src/github.com/agl/ed25519 ∴ go test -bench .
PASS
BenchmarkKeyGeneration 20000 85880 ns/op
BenchmarkSigning 20000 89115 ns/op
BenchmarkVerification 10000 212585 ns/op
ok github.com/agl/ed25519 7.500s
BenchmarkKeyGeneration 30000 47007 ns/op
BenchmarkSigning 30000 48820 ns/op
BenchmarkVerification 10000 119701 ns/op
ok github.com/agl/ed25519 5.775s
Making key generation, signing, and verification a rough average of 33%
faster, 44% faster, and 43% faster respectively. Of course, this
Making key generation and signing a rough average of 2x faster, and
verification 2.5-3x faster depending on the availability of avx2. Of course, this
is just my machine, and these results—nowhere near rigorous—should be taken
with a handful of salt.
Translating to a rough cycle count: we multiply by a factor of 2.6 to convert
nanoseconds to cycles per second on a 2591 Mhz CPU, that's 237660 cycles for
verification and 102523 for signing, which for signing is competitive
with optimised assembly versions.
Translating to a rough cycle count: we multiply by a factor of 3.3 to convert
nanoseconds to cycles per second on a 3300 Mhz CPU, that's 110256 cycles for
verification and 52618 for signing, which is competitive with hand-optimised
assembly implementations.
Additionally, if you're using a CSPRNG from the `rand` crate, the `nightly`
feature will enable `u128`/`i128` features there, resulting in potentially