Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 20 Jan 2012 21:35:26 +0000
From: Alex Sicamiotis <alekshs@...mail.com>
To: <john-users@...ts.openwall.com>
Subject: RE: DES with OpenMP



> Date: Sat, 21 Jan 2012 00:52:17 +0400
> From: solar@...nwall.com
> To: john-users@...ts.openwall.com
> Subject: Re: [john-users] DES with OpenMP
>
> On Fri, Jan 20, 2012 at 08:30:53PM +0000, Alex Sicamiotis wrote:
> > anyway gcc 4.6/4.7 was crap
>
> Indeed. Please try 4.3. Really.
>

I'll check it out.

> That's puzzling. Frankly, I am not convinced that your analysis is
> correct, although it might be.
>

These are 2 builds with icc... one with -fopenmp and one with no openMP (and ASM=0 for equality). As for flags, instead of -O2, I've used -fast and -march=core2.

CPU running at 3.66 GHz / test under KDE desktop

linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-icc-OMP-fast-marchcore2 -test     
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     7849K c/s real, 3984K c/s virtual
Only one salt:  5924K c/s real, 3019K c/s virtual

Benchmarking: BSDI DES (x725) [128/128 BS SSE2-16]... ^CWait...
Session aborted
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # export OMP_NUM_THREADS=1
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-icc-OMP-fast-marchcore2 -test
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     4235K c/s real, 4235K c/s virtual
Only one salt:  3858K c/s real, 3874K c/s virtual

Benchmarking: BSDI DES (x725) [128/128 BS SSE2-16]... ^CWait...
Session aborted
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-icc-noOMP-fast-marchcore2 -test
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     4020K c/s real, 4028K c/s virtual 
Only one salt:  3857K c/s real, 3857K c/s virtual

Benchmarking: BSDI DES (x725) [128/128 BS SSE2-16]... ^CWait...
Session aborted
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-icc-OMP-fast-marchcore2 -test
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     4238K c/s real, 4247K c/s virtual 
Only one salt:  3901K c/s real, 3901K c/s virtual

Benchmarking: BSDI DES (x725) [128/128 BS SSE2-16]... ^CWait...
Session aborted
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-icc-noOMP-fast-marchcore2 -test
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     4057K c/s real, 4065K c/s virtual
Only one salt:  3888K c/s real, 3888K c/s virtual


...the OMP version is scoring +200k in many salts over the noOMP despite threads=1, when it should be equal. This implies a boost from the second core (and top confirms this boost). Note that this is with icc and the intel openmp library. The gap on GCC is even larger (fresh GCC build / -O2 -march=nocona / ASM=0) :



linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # export OMP_NUM_THREADS=2
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-GCC-O2-nocona-OMP -test (dual thread performance)
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     7069K c/s real, 3610K c/s virtual
Only one salt:  5626K c/s real, 2873K c/s virtual

Benchmarking: BSDI DES (x725) [128/128 BS SSE2-16]... ^CWait...
Session aborted
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # export OMP_NUM_THREADS=1
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-GCC-O2-nocona-OMP -test
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     3803K c/s real, 3803K c/s virtual
Only one salt:  3495K c/s real, 3502K c/s virtual

Benchmarking: BSDI DES (x725) [128/128 BS SSE2-16]... ^CWait...
Session aborted
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-GCC-O2-nocona-noOMP -test
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     3098K c/s real, 3104K c/s virtual  (******noOMP = -700k over OMP threads=1****)
Only one salt:  3004K c/s real, 3004K c/s virtual

Benchmarking: BSDI DES (x725) [128/128 BS SSE2-16]... ^CWait...
Session aborted
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-GCC-O2-nocona-OMP -test
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     3753K c/s real, 3753K c/s virtual
Only one salt:  3440K c/s real, 3447K c/s virtual

Benchmarking: BSDI DES (x725) [128/128 BS SSE2-16]... ^CWait...
Session aborted
linux-1mo8:~/Documents/john-1.7.9icc/john-1.7.9/run # ./john-GCC-O2-nocona-noOMP -test
Benchmarking: Traditional DES [128/128 BS SSE2-16]... DONE
Many salts:     3076K c/s real, 3082K c/s virtual (******noOMP = -700k over OMP threads=1****)
Only one salt:  2973K c/s real, 2979K c/s virtual

These extra 700/400k (many/one salt) are from the second cpu thread despite the threads=1.
 		 	   		  

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.