[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 8 Jul 2012 10:50:20 +0200
From: Frank Dittrich <frank_dittrich@...mail.com>
To: john-dev@...ts.openwall.com
Subject: Re: -atom make targets
On 07/08/2012 08:44 AM, Solar Designer wrote:
> On Sat, Jul 07, 2012 at 11:49:48AM +0200, Frank Dittrich wrote:
>> Benchmarking: FreeBSD MD5 [32/32]... 3157 c/s real, 3157 c/s virtual
>> Benchmarking: FreeBSD MD5 [32/32 X2]... 2720 c/s real, 2720 c/s virtual
>> Benchmarking: FreeBSD MD5 [32/32]... 2360 c/s real, 2360 c/s virtual
>
> I think this suggests that MD5_IMM = 0 is a better choice for Atom.
> This is totally unexpected to me. MD5_IMM = 0 is normally for RISC
> archs, not for x86.
Repeating the generic build with latest git, I get:
$ make clean generic > /dev/null
Benchmarking: Traditional DES [24/32 4K]... 57600 c/s real, 57600 c/s
virtual
Benchmarking: Traditional DES [24/32 128K]... 51456 c/s real, 51456 c/s
virtual
Benchmarking: Traditional DES [24/32 4K]... 58240 c/s real, 58828 c/s
virtual
Benchmarking: Traditional DES [24/32 4K]... 74880 c/s real, 74880 c/s
virtual
Benchmarking: Traditional DES [24/32 4K]... 70784 c/s real, 70784 c/s
virtual
Benchmarking: Traditional DES [32/32 BS]... 170112 c/s real, 170112 c/s
virtual
Benchmarking: Traditional DES [32/32 BS]... 180416 c/s real, 180416 c/s
virtual
Benchmarking: Traditional DES [32/32 BS]... 106752 c/s real, 106752 c/s
virtual
Benchmarking: FreeBSD MD5 [32/32]... 3160 c/s real, 3160 c/s virtual
Benchmarking: FreeBSD MD5 [32/32 X2]... 2700 c/s real, 2700 c/s virtual
Benchmarking: FreeBSD MD5 [32/32]... 3193 c/s real, 3193 c/s virtual
Benchmarking: OpenBSD Blowfish (x32) [32/32]... 184 c/s real, 186 c/s
virtual
Benchmarking: OpenBSD Blowfish (x32) [32/32]... 220 c/s real, 220 c/s
virtual
Benchmarking: OpenBSD Blowfish (x32) [32/32 X2]... 170 c/s real, 170 c/s
virtual
So, this time the 3rd MD5 value is similar to the 1st.
May be last time I unplugged the power supply in the middle of the
build, I don't know.
c/s rate in benchmark for generic build is as disappointing as it was
for the 1.7.9 core version:
Benchmarking: FreeBSD MD5 [32/32]... DONE
Raw: 3191 c/s real, 3197 c/s virtual
This is for linux-x86-any:
Benchmarking: FreeBSD MD5 [32/32]... DONE
Raw: 4252 c/s real, 4260 c/s virtual
And this for linux-x86-sse2:
Benchmarking: FreeBSD MD5 [128/128 SSE2 intrinsics 12x]... DONE
Raw: 6524 c/s real, 6537 c/s virtual
So, in current git, sse2 is much faster than any, and generic fails
again to provide a reasonable performance.
Frank
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.
