This imports the crypto/ed25519/internal/edwards25519/field package from
CL 276272, and uses it in x/crypto/curve25519.
The ScalarMult code was ported 1:1 from curve25519_generic.go.
old code lines new code lines
Go 896 463
Assembly (manually written) 1772 (1772) 362 (34)
43% performance loss on amd64, 33% loss on 386, and 45% gain on arm64.
Feels worth it to remove 1700 lines of manually written assembly.
Apple M1
name old time/op new time/op delta
X25519Basepoint-8 85.0µs ± 1% 46.4µs ± 0% -45.39% (p=0.000 n=10+9)
X25519-8 84.4µs ± 0% 46.7µs ± 2% -44.76% (p=0.000 n=8+9)
Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz
name old time/op new time/op delta
X25519Basepoint-4 42.6µs ± 1% 60.9µs ± 1% +43.22% (p=0.000 n=9+10)
X25519-4 42.5µs ± 1% 60.9µs ± 0% +43.17% (p=0.000 n=9+9)
Intel(R) Core(TM) i5-7400 CPU @ 3.00GHz [GOARCH=386]
name old time/op new time/op delta
X25519Basepoint-4 530µs ± 1% 703µs ± 1% +32.81% (p=0.000 n=10+10)
X25519-4 530µs ± 1% 706µs ± 1% +33.18% (p=0.000 n=10+10)
Change-Id: I1dc62a6a3a3e417a1366ff873c475087a0395124
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/315269
Run-TryBot: Filippo Valsorda <filippo@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Filippo Valsorda <filippo@golang.org>
Trust: Katie Hockman <katie@golang.org>
Reviewed-by: Katie Hockman <katie@golang.org>
Comparison against BoringSSL-generated test vectors showed mismatches
with the pure Go implementation of curve25519. The problem was narrowed
down to a missing mask in feFromBytes(). This diff adds the mask,
bringing this back in line with the reference implementation and
RFC 7748:
When receiving such an array, implementations of X25519 (but not
X448) MUST mask the most significant bit in the final byte. This is
done to preserve compatibility with point formats that reserve the
sign bit for use in other protocols and to increase resistance to
implementation fingerprinting.
Fixesgolang/go#30095
Change-Id: If7efc0e2acd6efb761d6e3cb89cec359d7d81cb1
Reviewed-on: https://go-review.googlesource.com/c/161257
Run-TryBot: Filippo Valsorda <filippo@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Simplify the constant swap function.
On amd64: Replace the CMOVQEQ scheme with SSE2 code similar to the non-amd64 code.
On non-amd64: Avoid unnecessary loop iterations.
The result is less and slightly faster code.
name old time/op new time/op delta
ScalarBaseMult-4 653µs ± 0% 636µs ± 0% ~ (p=0.100 n=3+3)
name old time/op new time/op delta
ConstantSwap-4 10.4ns ± 1% 6.2ns ± 0% -39.86% (p=0.029 n=4+4)
On an i7-65000U
Change-Id: Ia5eea92e0b3eabb6c291d25229aa582b51278552
Reviewed-on: https://go-review.googlesource.com/39693
Reviewed-by: Adam Langley <agl@golang.org>
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Previously curve25519 contained a constant-time, optimised amd64 implementation and
a generic implemenation that used math/big and that was not constant-time.
This change contains a Go port of the public domain, "ref10" implementation from
SUPERCOP. This has the advantage of being faster and constant-time.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/13343045
This consists of ~2000 lines of amd64 assembly and a, much slower,
generic Go version in curve25519.go. The assembly has been ported from
djb's public domain sources and the only semantic alterations are to
deal with Go's split stacks.
R=rsc
CC=golang-dev
https://golang.org/cl/5786045