Re: [Gems-users] Transactional MicroBenchmark

[bobba@xxxxxxxxxxx]

There is no mechanism in the dequeue benchmark to bind threads to
processors. If you wish to do so, you need to add processor_bind calls as
indicated in the example code.

Jayaram

> Hi All;
> 	I am trying to understand the code of the transactional benchmark --
> dequeue, and the mechanism in dequeue that binds a thread to a
> processor is not clear to me. It seems in the code that the pthread
> attr is set to SYSTEM_SCOPE rather than the default processor scope
> and that's all that makes the pthread bind to a processor.
> 	By looking around on the Web, in  HW2 of cs 838 by Prof. Mark Hill,
> the example code shows that there is an explicit call for binding
> pthread to processor as the following:
>    /* ********************** Thread Initialization
> *********************** */
>    /* Bind the thread to a processor.  This will make sure that each of
>     * threads are on a different processor.  ProcessorIds[threadId]
>     * specifies the processor ID which the thread is binding to.
>     */
>    ret = processor_bind(P_LWPID, P_MYID, ProcessorIds[threadId], NULL);
>
> But the example code uses a different compiler than the dequeue
> benchmark, could this be the cause or there is something missing in
> the dequeue benchmark?
>
> The example code can be found on the following link:
> http://www.cs.wisc.edu/~markhill/cs838/Fall2005/handouts/eg_pthread.zip
>
> Some explaination for the example code can be found on the following
> link:
> http://www.cs.wisc.edu/~markhill/cs838/Fall2005/handouts/hw2.html
>
> 	Thanks for any reply in advance.
>
> Regards
> James
>
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
>