Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 16 Jan 2012 16:48:46 +0100
From: magnum <>
Subject: Re: SHA-crypt

On 01/16/2012 10:53 AM, magnum wrote:
> In progress. First incarnation was 35% faster than generic-crypt format
> already (using same test vectors) although my only goal was for it to
> work at all.

Pushed to GitHub now, both sha-crypt256 and sha-crypt512. I didn't find
much to optimise, it's spending over 95% in the OpenSSL functions. But
OMP should scale well (I only tested two cores).

Benchmarking: generic crypt(3) SHA-256 rounds=5000 [?/64]... (2xOMP) DONE
Many salts:	372 c/s real, 189 c/s virtual
Only one salt:	372 c/s real, 190 c/s virtual

Benchmarking: SHA-crypt-256 [OpenSSL 32/64]... (2xOMP) DONE
Many salts:	497 c/s real, 252 c/s virtual
Only one salt:	497 c/s real, 254 c/s virtual

Benchmarking: generic crypt(3) SHA-512 rounds=5000 [?/64]... (2xOMP) DONE
Many salts:	432 c/s real, 221 c/s virtual
Only one salt:	436 c/s real, 220 c/s virtual

Benchmarking: SHA-crypt-512 [OpenSSL 64/64]... (2xOMP) DONE
Many salts:	579 c/s real, 292 c/s virtual
Only one salt:	579 c/s real, 292 c/s virtual

crypt_all() and rounds parsing are pretty close copies of the reference
code. Self-tests are the same that existed in crypt_fmt.c. I stole a
little base64 convenience from MD5_std.c too.


Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.