This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Benchmark for sem_timedwait v1.1
- From: Siddhesh Poyarekar <siddhesh dot poyarekar at gmail dot com>
- To: Ondřej Bílka <neleai at seznam dot cz>
- Cc: Roland McGrath <roland at hack dot frob dot com>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Wed, 26 Feb 2014 19:44:12 +0530
- Subject: Re: Benchmark for sem_timedwait v1.1
- Authentication-results: sourceware.org; auth=none
- References: <20140211172720 dot 52ABB7445F at topped-with-meat dot com> <20140211184530 dot GB32451 at domone dot podge> <20140211204751 dot GA17236 at domone dot podge> <20140213102200 dot GA8107 at spoyarek dot pnq dot redhat dot com> <20140216152649 dot GA10444 at domone dot podge> <20140217050742 dot GC17663 at spoyarek dot pnq dot redhat dot com> <20140217101900 dot GA16967 at domone dot podge> <20140217103551 dot GF27461 at spoyarek dot pnq dot redhat dot com> <20140221192656 dot GB6229 at domone dot podge> <20140222050816 dot GA9254 at spoyarek dot pnq dot redhat dot com> <20140226135803 dot GD22604 at domone dot podge>
On 26 February 2014 19:28, OndÅej BÃlka <neleai@seznam.cz> wrote:
> Its artificial, you could add always add additional parameter, or for
> general benchmark take a string as argument, then call system on it and
> require that string be something like "echo 'code' > test.c; gcc test.c; ./a.out"
It is not artificial and it is definitely not something that the
framework was not intended to do. I've hacked up a quite example that
does exactly what your test does, but within the framework - see the
branch siddhesh/sem_timedwait.
It needs one additional feature to the framework, which is the ability
to define an initialization function; something that I should have
added ages ago. Other than that it has exactly what i had mentioned -
a benchmark input file that has two variants (which is your
requirement) to the scenario being benchmarked and a source file that
has a function with the initializer and the scenario.
I've not posted these patches because I'm waiting for Mike's review of
the benchmark conversion to python. I will post them once that is
done.
>> This particular benchmark you posted does not spawn threads; it only
>> calls a set of functions. I agree that in most cases (not necessarily
>> all though) that involve spawning threads and measuring events across
>> threads, bench-skeleton would be a bad fit.
>>
> As that would eventually happens and even now there is notrivial code
> for setup there is no point in trying to fix a
> round cube to square holes.
I don't understand what you're trying to say.
Siddhesh
--
http://siddhesh.in