fix: align all 3 benchmarks to identical operation set
All 3 platform benchmarks (Android, JVM, K/Native) now measure the
same core operations for cross-platform comparability:
verifySchnorr — strict BIP-340
verifySchnorrFast — Nostr x-check only (no y-parity inversion)
signSchnorr — derives pubkey each time
signSchnorr(cached) — pre-computed x-only pubkey
compressedPubKeyFor — create + compress
secKeyVerify — key validation
privKeyTweakAdd — BIP-32 key derivation
ecdhXOnly — Nostr ECDH (NIP-04/44)
batch verify (8, 16) — same-pubkey batch
Changes:
- Android: added signSchnorrCachedPkOurs, ecdhXOnly/ecdhXOnlyOurs
(replaced pubKeyTweakMulCompact naming)
- JVM: removed pubkeyCreate and pubKeyCompress (subsets of
compressedPubKeyFor, not measured on other platforms)
- K/Native: added verifySchnorrFast
K/Native retains field micro-benchmarks (FieldP.mul/sqr/add/sub/inv,
unsignedMultiplyHigh, uLt) as a K/N-specific diagnostic tool.
https://claude.ai/code/session_01EMY5RnXb9rnsyU2KbXrSaY
Claude
da64a4a4d3