Performance vs hashlib (SHA-512)

[bschollnick]

I was hoping to see a speed increase from hashlib, but pycryptodome appears to be running ~6x slower that hashlib with SHA-512 generation...

python 3.7.1 (64bit) on windows 10 64-bit.

`from random import randint
import string
import timeit
import hashlib

builtin = """
import hashlib
import random
import string
max = 1024
randomString = "This is a test only a test, of the python string, going through this shite, to see if there is any performance difference between hashlib and a different library"
sha = hashlib.sha512()
sha.update(randomString.encode("utf-16"))
sha.hexdigest()
"""

pcrypto = """
from Crypto.Hash import SHA512
import random
import string
max = 1024
randomString = "This is a test only a test, of the python string, going through this shite, to see if there is any performance difference between hashlib and a different library"
sha = SHA512.new()
sha.update(randomString.encode("utf-16"))
sha.hexdigest()
"""
print ("hashlib - ",timeit.timeit(stmt=builtin, number=10000))
print ("PCrytpo - ",timeit.timeit(stmt=pcrypto, number=10000))
`
hashlib - 0.036552100000000004
PCrytpo - 0.4269634

Is this expected? I'm happy to work with you to diagnose this further...
I have code that is heavily utilizing SHA-512 hashes and if there is a faster way to generate them, I'm more than happy to put some effort in to assist.

[Legrandin]

hashlib is backed by openssl, which implements SHA512 in assembly using AVX instructions (assuming you run a modern X86 CPU), so yes it is expected to be faster than PyCryptodome, which only uses portable C (with some SSE2 intrinsics here and there).