Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130418230345.C38QG.37120.imail@eastrmwml206>
Date: Thu, 18 Apr 2013 23:03:45 -0400
From:  <jfoug@....net>
To: john-dev@...ts.openwall.com
Subject: Re: Got all dyna formats (except $1$ and $apr1$) working
 with OMP

---- magnum <john.magnum@...hmail.com> wrote: 
> This is without looking at any code, but could you not address this simply by bumping BLOCK_LOOPS (or equivalent) by 10 (perhaps 16) for OMP? So if a non-OMP build does 10 of them, OMP would do 160. Would that mitigate the impact you describe?

Not that easy.

I did make a quick prototype change.  I change the increment value from what was computed per type, to 1440.  This would make a OMP_NUM_THREADS run in only 4 loops.

This made 1 thread OMP work almost same speed as non-OMP, for 'some' dynas.  However, in others, things were bad.  60%, 50% and even some slower than that (40% or so).

I THINK this is due to unicode checking, calling omp_thread_num() within many of the string functions.

I will now try to build the code to take 3 params (3 ints).  So it will be start, stop and threadID as params.  that way, the thread id will be available as a param, when needed.  This might slow down formats overall a percent or so, but should remove that OMP overhead on a lot of the string appending functions, IF that is the problem.

Jim.

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.