crypto/elliptic,crypto/ecdsa: P256 amd64 assembly
This is based on the implementation used in OpenSSL, from a
submission by Shay Gueron and myself. Besides using assembly,
this implementation employs several optimizations described in:
S.Gueron and V.Krasnov, "Fast prime field elliptic-curve
cryptography with 256-bit primes"
In addition a new and improved modular inverse modulo N is
implemented here.
The performance measured on a Haswell based Macbook Pro shows 21X
speedup for the sign and 9X for the verify operations.
The operation BaseMult is 30X faster (and the Diffie-Hellman/ECDSA
key generation that use it are sped up as well).
The adaptation to Go with the help of Filippo Valsorda
Updated the submission for faster verify/ecdh, fixed some asm syntax
and API problems and added benchmarks.
Change-Id: I86a33636747d5c92f15e0c8344caa2e7e07e0028
Reviewed-on: https://go-review.googlesource.com/8968
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>