linux-kernel - Some results (was: Re: [RFC PATCH] x86, fpu: Use eagerfpu by default on all CPUs)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20150225171227.GJ3226@pd.tnic>
Date:	Wed, 25 Feb 2015 18:12:27 +0100
From:	Borislav Petkov <bp@...en8.de>
To:	Andy Lutomirski <luto@...capital.net>
Cc:	Oleg Nesterov <oleg@...hat.com>, Rik van Riel <riel@...hat.com>,
	x86@...nel.org, linux-kernel@...r.kernel.org,
	Mel Gorman <mgorman@...e.com>
Subject: Some results (was: Re: [RFC PATCH] x86, fpu: Use eagerfpu by default
 on all CPUs)

On Fri, Feb 20, 2015 at 10:58:15AM -0800, Andy Lutomirski wrote:
> -	/* Auto enable eagerfpu for xsaveopt */
> -	if (cpu_has_xsaveopt && eagerfpu != DISABLE)
> +	/* Auto enable eagerfpu for everyone */
> +	if (eagerfpu != DISABLE)
>  		eagerfpu = ENABLE;

So Mel did run some measurements with it on an old Intel and AMD box.
Attached are netperf results from both. The Intel box is a quad core
very similar to this one:

http://ark.intel.com/products/28020/Intel-Xeon-Processor-5063-4M-Cache-3_20-GHz-1066-MHz-FSB

netperf-udp-rr shows some impact of the eagerfpu patch which is outside
of the noise level. netperf-tcp-rr not so much but eager is still a bit
behind.

The AMD box is an old K8 and results there look like eager is better :-)
Not with all though - pipetest is worse.

netperf-udp-rr is better at almost every data point and tcp-rr looks a
bit useless with those high noise levels.

As a summary, the patch has some, albeit small, impact. We would need
more benchmarks. We have one speccpu run which is currently taking
forever to finish...

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

Download attachment "amd.tar.bz2" of type "application/octet-stream" (19665 bytes)

Download attachment "intel.tar.bz2" of type "application/octet-stream" (19099 bytes)