Closed Sc00bz closed 11 months ago
Thank you for the PR and sorry about the delay!
The change is correct and has a bit of performance effect on the micro-benchmark, but in my opinion not enough to justify diverting from the comment above the code, for readability reasons.
go: go1.21.4
goos: darwin
goarch: arm64
pkg: filippo.io/edwards25519
│ v1.1.0 │ v1.1.0-1-gbd55154c304 │
│ sec/op │ sec/op vs base │
ScalarInversion-8 11.77µ ± 0% 11.73µ ± 0% -0.37% (p=0.000 n=25)
Yeah, no problem.
Instead of using
table[1/2]
which ist
usett
which ist**2
.