[parsec-users] Profiling PARSEC BENCHMARKS

Anusha anu.jeya at gmail.com
Sat Nov 20 16:06:58 EST 2010


I am mainly interested in finding out the overhead of POSIX primitives, in
the sense, the time a thread spends waiting for conditionals, barriers and
mutexes.

I will modify the config file and test with gprof enabled and provide an
update here.

-Anusha

2010/11/20 Christian Bienia <cbienia at cs.princeton.edu>

> Anusha,
>
>
>
> What profiling strategy is the best depends on what type of information
> you’re interested in. Check the output of the profiling tools to see what it
> does and make sure it gives you what you’re looking for.
>
>
>
> You can enable gprof support for the PARSEC benchmarks by adapting the
> build configuration files. How to do that is described in the PARSEC
> tutorial available from the PARSEC website.
>
>
>
> Best,
> Chris
>
>
>
>
>
> *From:* parsec-users-bounces at lists.cs.princeton.edu [mailto:
> parsec-users-bounces at lists.cs.princeton.edu] *On Behalf Of *Anusha
> *Sent:* Saturday, November 20, 2010 3:22 PM
> *To:* parsec-users at lists.cs.princeton.edu
> *Subject:* [parsec-users] Profiling PARSEC BENCHMARKS
>
>
>
> Hi,
>
>
>
> I am planning to profile the PARSEC benchmarks for my project. I plan to
> profile the pthreads.
>
>
>
> The ideas I came up were
>
>
>
> -Instrumenting the glibc.
>
>
>
> -Using gprof. But I am not sure how to enable gprof while building parsec.
> Is this possible? I actually have a work around to use gprof
> for multi-threaded application but I don't know how to enable gprof while
> building parse.
>
>
>
> Is there are any other better method to profile the benchmark or one
> amongst the two would work?
>
>
>
> -Anusha
>
>
>
> _______________________________________________
> parsec-users mailing list
> parsec-users at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/parsec-users/attachments/20101120/cee1ca05/attachment.html>


More information about the parsec-users mailing list