Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Mon, 24 Aug 2015 12:38:03 +0300
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: md5crypt-opencl

magnum, all -

I've experimented with optimizing our md5crypt-opencl in various ways -
patch attached.

Before:

Device 1: Tahiti [AMD Radeon HD 7900 Series]
Benchmarking: md5crypt-opencl, crypt(3) $1$ [MD5 OpenCL]... DONE
Raw:    1966K c/s real, 28086K c/s virtual

0g 0:00:00:37 0.00% (ETA: 2030-08-03 18:33) 0g/s 435.9p/s 2839Kc/s 2839KC/s egyaaaaa..hmwbaaaa

Device 5: GeForce GTX TITAN
Benchmarking: md5crypt-opencl, crypt(3) $1$ [MD5 OpenCL]... DONE
Raw:    1512K c/s real, 1527K c/s virtual

0g 0:00:00:37  0g/s 0p/s 1876Kc/s 1876KC/s GPU:61б╟C util:94% fan:39% aaaaaaaa..xovdbaaa

After:

Device 1: Tahiti [AMD Radeon HD 7900 Series]
Benchmarking: md5crypt-opencl, crypt(3) $1$ [MD5 OpenCL]... DONE
Raw:    2146K c/s real, 23847K c/s virtual

0g 0:00:00:37 0.00% (ETA: 2030-08-03 18:40) 0g/s 434.8p/s 2988Kc/s 2988KC/s egyaaaaa..hmwbaaaa

or sometimes:

Raw:    2162K c/s real, 24029K c/s virtual

0g 0:00:00:37 0.00% (ETA: 2030-08-03 18:44) 0g/s 434.8p/s 2992Kc/s 2992KC/s egyaaaaa..hmwbaaaa

and sometimes it gets up to ~3005K.

Device 5: GeForce GTX TITAN
Benchmarking: md5crypt-opencl, crypt(3) $1$ [MD5 OpenCL]... DONE
Raw:    1638K c/s real, 1638K c/s virtual

0g 0:00:00:37  0g/s 0p/s 1861Kc/s 1861KC/s GPU:62б╟C util:99% fan:39% aaaaaaaa..xovdbaaa

With the inline ptx asm enabled (need proper compile-time check for it):

Raw:    1664K c/s real, 1664K c/s virtual

0g 0:00:00:37  0g/s 0p/s 1875Kc/s 1875KC/s GPU:62б╟C util:86% fan:39% aaaaaaaa..xovdbaaa

The actual cracking benchmarks here are for length 8 - possible changes
in handling of odd lengths are not seen there.  I also ran length 7 and
9 benchmarks to see those.

The --test benchmarks run our usual mix of lengths, which is why they
are much worse.

As discussed in another thread, we probably need to try uint2 for this
kernel on Kepler.

Alexander

View attachment "john-md5crypt-opencl-opt1.diff" of type "text/plain" (10251 bytes)

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.