[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 3 Sep 2020 22:14:58 +0200
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: SIMD performance impact
On Tue, Sep 01, 2020 at 11:48:02AM +0200, Vincent wrote:
> For sure it's possible. So I would be interested in a:
>
> cd $john/src/
> lscpu
> for simd in sse4.2 avx avx2 avx512f avx512bw
> do
> ./configure --enable-simd=$simd
> make -s clean
> make -s
> echo "$simd"
> ../run/john --test
> done
This is specific enough that I might just run something similar on a
couple of machines and post the results later.
> I just haven't got the latest and greatest AMD and Intel CPUs to run it
> myself ;)
In my very limited experience with recent AMD CPUs, for AMD Zen (first
generation Epyc) the results are very similar to those for Intel
Broadwell (such as in E5-26xx v4). In fact, so similar that the same
code optimizations apply to both, which is convenient. This includes
AVX2 for formats where we support it.
For newer CPUs, it's trickier: AMD got more cores (and from the recent
bcrypt benchmark I posted, apparently also more scalar execution units
per core), whereas Intel got AVX-512.
Where AVX-512 is supported, it typically provides a speedup of somewhat
below 2x vs. AVX2 on the same CPU, because the clock rate and cache hit
rate are lower. Can be e.g. 1.8x. For memory-bound algorithms like
Argon2, the speedup is smaller or non-existent (it's substantial when
running few threads, but diminishes when approaching the hardware's
maximum), although these are rare in password cracking so far.
> Central documentation would be a big plus for others I guess, for future
> users with similar questions.
Yeah. The sort of benchmark results we currently have on the wiki are
mostly a historical curiosity, and ideally we should come up with some
currently relevant replacement.
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.
