|
Message-ID: <20130121154155.GA11300@openwall.com> Date: Mon, 21 Jan 2013 19:41:55 +0400 From: Solar Designer <solar@...nwall.com> To: john-dev@...ts.openwall.com Subject: Re: Proposed optimizations to pwsafe On Sun, Jan 20, 2013 at 11:14:55PM -0500, Brian Wallace wrote: > I got some time before I thought I would have any, and dug around the > source to get an idea of how data flows in JtR. I've done a preliminary > implementation of the optimization in my forked repository, with some gains. Thank you! > Old: > ./john --test=15 --format=pwsafe > Benchmarking: Password Safe SHA-256 [32/64]... (6xOMP) DONE > Raw: 7027 c/s real, 1213 c/s virtual > > New: > ./john --test=15 --format=pwsafe > Benchmarking: Password Safe SHA-256 [32/64 OpenSSL]... (6xOMP) DONE > Raw: 8268 c/s real, 1395 c/s virtual I've just tested on FX-8120, and I am seeing a similar difference: Old: Benchmarking: Password Safe SHA-256 [32/64 OpenSSL]... (8xOMP) DONE Raw: 6857 c/s real, 854 c/s virtual New: Benchmarking: Password Safe SHA-256 [32/64 OpenSSL]... (8xOMP) DONE Raw: 8309 c/s real, 1041 c/s virtual BTW, the " OpenSSL" comment is now wrong and should be removed. Meanwhile, here's a tweet by atom: <hashcat> Added kernel for cracking "Passwords Safe" container to hashcat and oclHashcat! Its faster than one would expect: hd7970: 501k, FX8120: 24k 24k is great, but not surprising - it only means that atom actually has SIMD code for SHA-256, unlike us. We need to implement it. At this time, we have SIMD code for MD4, MD5, SHA-1, but not yet for SHA-256 and SHA-512. We have an unclaimed task here: http://openwall.info/wiki/ideas#Tasks1 (OK, the description is out of date, now that hashcat clearly has SIMD implementation of at least SHA-256, and perhaps also SHA-512.) As to HD 7970, here's our current speed: $ ./john -te -fo=pwsafe-opencl -pla=1 OpenCL platform 1: AMD Accelerated Parallel Processing, 2 device(s). Device 0: Tahiti (AMD Radeon HD 7900 Series) Local worksize (LWS) 64, Global worksize (GWS) 57344 Benchmarking: Password Safe SHA-256 [OpenCL]... DONE Raw: 353975 c/s real, 17203K c/s virtual hashcat's is better, too (354k vs. 501k). Lukas? Alexander
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.