This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] benchtests/Makefile: Run the string benchmarks four times by default.
- From: Will Newton <will dot newton at linaro dot org>
- To: Siddhesh Poyarekar <siddhesh at redhat dot com>
- Cc: libc-alpha <libc-alpha at sourceware dot org>, Patch Tracking <patches at linaro dot org>
- Date: Thu, 5 Sep 2013 11:36:00 +0100
- Subject: Re: [PATCH] benchtests/Makefile: Run the string benchmarks four times by default.
- Authentication-results: sourceware.org; auth=none
- References: <52274838 dot 7010902 at linaro dot org> <20130905103119 dot GJ4306 at spoyarek dot pnq dot redhat dot com>
On 5 September 2013 11:31, Siddhesh Poyarekar <siddhesh@redhat.com> wrote:
> On Wed, Sep 04, 2013 at 03:48:24PM +0100, Will Newton wrote:
>>
>> The string benchmarks can be affected by physical page placement, so
>> running them multiple times is required to account for this. Also
>> backup the results of the previous run like is done for the other
>> benchmarks.
>
> I wonder if this would make a lot of difference. On an otherwise idle
> system (which is essentially a pre-condition for benchmark runs), I
> would expect more or less similar page placement (i.e. reachable via
> comparable indices) for consecutive executions of a program. You
> would probably need to have something spawned in parallel to actually
> make a difference.
There is no guarantee the OS will give you a different set of physical
pages but it does seem to have the desired effect in practice.
--
Will Newton
Toolchain Working Group, Linaro