Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 3 Mar 2021 12:17:33 +0100
From: Solar Designer <>
Subject: Re: Splitting mask keyspace

On Wed, Mar 03, 2021 at 11:49:15AM +0100, Micha?? Majchrowicz wrote:
> Sorry for repost but I thought it's our previous conversation :D
> I have some results from testing --fork option. On a test node it was
> working pretty stable but to be honest I don't like the idea of
> "overbooking forks". It does kinda work but has limited functionality.
> If I have see2, avx and avx2 nodes with same number of cores I have to
> use 8, 16 and 32 in order to correctly "overbook". The issue is that
> on some of multi core machines I don't want to use all cores for john.
> I understand why it's designed like this but since I can already
> overbook forks it would be nice to have an option to ONLY overbook
> node option so it would be divisible by number of forks. I think
> adding this would be much simpler than splitting the correlation
> between fork and node and would solve my issue :)

Now this is a feature request, and you've correctly already created a
GitHub issue for it:

"Allow fork divisible node values"

My reply:

"What I was thinking is that we should allow e.g. --fork=32
--node=200-263/333, which would run only 32 processes yet use up 64 node
numbers (each process would correspond to two virtual nodes, taking
their portions of the keyspace)."

If I understood you correctly, then I agree it's something we'd want to
implement (but I don't know who will do it nor when).

Let's continue this sub-thread on the GitHub issue only.


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.