[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 18 Dec 2011 20:15:22 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: 1.7.9-jumbo
On Sun, Dec 18, 2011 at 05:03:50PM +0100, magnum wrote:
> The change is that in 1.7.8, we always had 16K keys_per_crypt, and in
> 1.7.9 we have 8K x number of threads (the latter is capped to 4, because
> it does not scale further). I did that on some AMD hardware using
> gcc-4.5 and I did not see this performance drop (on the contrary).
>
> The fastest formats will always show these variations depending on
> hardware and exact compiler version but we could possibly change this to
> 16K x number of threads and see if it helps.
This makes sense to me.
Here's what I get on 2xE5420. 1.7.8-jumbo-8:
Benchmarking: CRC-32 [32/64]... (8xOMP) DONE
Many salts: 247758K c/s real, 30931K c/s virtual
Only one salt: 26427K c/s real, 3299K c/s virtual
Benchmarking: CRC-32 [32/64]... (8xOMP) DONE
Many salts: 249135K c/s real, 31141K c/s virtual
Only one salt: 26411K c/s real, 3289K c/s virtual
1.7.9-jumbo-5:
Benchmarking: CRC-32 [32/64]... (4xOMP) DONE
Many salts: 196575K c/s real, 49021K c/s virtual
Only one salt: 22347K c/s real, 5586K c/s virtual
Benchmarking: CRC-32 [32/64]... (4xOMP) DONE
Many salts: 207585K c/s real, 51766K c/s virtual
Only one salt: 33751K c/s real, 8416K c/s virtual
I haven't tried making changes yet.
Number of benchmarks: 153
Minimum: 0.79342 real, 0.56158 virtual
Maximum: 12.03732 real, 8.47542 virtual
Median: 1.00202 real, 1.00271 virtual
Median absolute deviation: 0.01500 real, 0.01710 virtual
Geometric mean: 1.16469 real, 1.06979 virtual
Geometric standard deviation: 1.57763 real, 1.35698 virtual
Number of benchmarks: 153
Minimum: 0.83322 real, 0.56479 virtual
Maximum: 12.04051 real, 8.46831 virtual
Median: 1.00429 real, 1.00097 virtual
Median absolute deviation: 0.01532 real, 0.01479 virtual
Geometric mean: 1.16748 real, 1.07191 virtual
Geometric standard deviation: 1.57355 real, 1.36328 virtual
(two runs of each version to be more certain of the results).
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.
