[parsec-users] Which PARSEC benchmark is good to test lock contention?

kishore kumar kishoreguptaos at gmail.com
Wed Nov 19 19:13:03 EST 2014


I measured lock contention on Solaris as it has several nice utilities
e..g, prstat(1). As you know, the amount lock contention depends on several
factors: "number of threads", "number of cores", "input", etc. I always run
PARSEC programs with native inputs. Depending on the number of threads, I
see that the program spends between 30% to 90% elapsed time in user space
locks/condition variables etc.

You may want to look at the following paper:
Thread Reinforcer: Dynamically Determining Number of Threads via OS Level
Monitoring

Thanks.

Best,
Kishore Pusukuri

On Wed, Nov 19, 2014 at 4:06 PM, Milind Chabbi <Milind.Chabbi at rice.edu>
wrote:

> Thanks Kishore. Is it possible to share more details of this application
> and lock contention? For example, what fraction of time it spends trying to
> enter and leave critical section?
>
> On Wed, Nov 19, 2014 at 5:57 PM, kishore kumar <kishoreguptaos at gmail.com>
> wrote:
>
>> streamcluster
>>
>> Best,
>> Kishore Pusukuri
>>
>> On Wed, Nov 19, 2014 at 3:54 PM, Milind Chabbi <Milind.Chabbi at rice.edu>
>> wrote:
>>
>>> Hello,
>>> I am looking for a multi-threaded application that has high contention
>>> for a lock. Does PARSEC suit have any representative applications with high
>>> lock contention? Is so, which application(s)?
>>>
>>> -Milind
>>>
>>> _______________________________________________
>>> parsec-users mailing list
>>> parsec-users at lists.cs.princeton.edu
>>> https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>>>
>>>
>>
>> _______________________________________________
>> parsec-users mailing list
>> parsec-users at lists.cs.princeton.edu
>> https://lists.cs.princeton.edu/mailman/listinfo/parsec-users
>>
>>
>
> _______________________________________________
> 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/20141119/76ef15ec/attachment.html>


More information about the parsec-users mailing list