Date: Thu, 5 Oct 2006 10:32:06 +0400 From: Solar Designer <solar@...nwall.com> To: john-users@...ts.openwall.com Subject: Re: Re: JTR and os X macintel On Thu, Oct 05, 2006 at 05:53:47AM +0000, -. -PhanTom-. - wrote: > Oh well, AMD has been known as the king of > Floating Point Operations in the past, they will be again... I guess This has nothing to do with JtR since it does not use floating point operations at all. > > > I was under the impression that the high speeds on Macs were due to the > > > G4 and G5 cpus and that the OS had very little to do with it...? > > > > Correct. > > Does this apply to Linux vs Windows as well? Yes - there should be almost no performance difference for JtR on Linux vs. Windows (except for Windows 9x/ME and older). > As far as I could tell, latest version of JTR would theoretically yield > higher speeds on linux64 on an AMD64 cpu compared to windows... I have to be guessing as to where you derived this conclusion from. We had a discussion in here some months ago where you wanted to make a 64-bit build of JtR on Windows - and that is not currently supported. So you could not see whether your CPU would yield better performance when running the x86-64 16-register SSE2 code vs. the plain x86 8-register SSE2 code. I suggested that you try on a Linux/x86-64 system. I did not imply that the 16-register version would be faster for you - although it might be. > However, I did test this and saw no higher speeds on Linux64 compared to > windows32.... could be related to compilation though..not sure... It is entirely possible that the 16-register version is not any faster on your CPU. In fact, it may even be a little bit slower, such as because of the extra instruction prefixes. Either way, the difference should be within 10% - anything beyond that would be abnormal. Please note that SSE is 128-bit (at least from the application's point of view) regardless of whether you build for 32-bit or 64-bit. -- Alexander Peslyak <solar at openwall.com> GPG key ID: 5B341F15 fp: B3FB 63F4 D7A3 BCCC 6F6E FC55 A2FC 027C 5B34 1F15 http://www.openwall.com - bringing security into open computing environments -- To unsubscribe, e-mail john-users-unsubscribe@...ts.openwall.com and reply to the automated confirmation request that will be sent to you.
Powered by blists - more mailing lists
Powered by Openwall GNU/*/Linux - Powered by OpenVZ