RTEMS 4.11 runs slower as RTEMS 4.88 on a monoprocessor?

[Joel Sherrill]

On Fri, May 27, 2016 at 10:00 AM, Thibaud Backenstrass <
thibaud.backenstrass at phelma.grenoble-inp.fr> wrote:

> Hello,
>
> I made a comparison of the timing performances of both RTEMS 4.8 and RTEMS
> 4.11 using the exact same application. Basically, the application creates
> and starts a few tasks, then stops them almost immediately. It also
> includes a timing observation service to know how much time each operation
> requires.
> Both RTEMS 4.8 and 4.11 are build with networking, POSIX API and Ada
> support. The application is built the same way and executed under the TSIM
> simulator for SPARC/LEON3, using only one processor.
>
> I noticed that the execution of my application under RTEMS 4.11 is at
> least 10% slower than under RTEMS 4.8.0 and I am wondering:
> - did the build options (and especially the optimization options) changed
> between RTEMS 4.8 and RTEMS 4.11?
> - is the gcc compiler 4.9.4 less “efficient” than the version 4.2.0 used
> to build RTEMS 4.8.0?
> - is RTEMS 4.11 slower due to the fact that is supports multiprocessor
> (SMP) and that the structures (semaphores, ...) are more time consuming,
> even on a monoprocessor?
>
>
All of those are possibilities but I would lean to the third option more.

Is this a C or Ada test?

Can you run the timing tests on both versions so we can see what
operations themselves have gotten slower?


> Or what could be the reasons why RTEMS 4.11 runs slower than RTEMS 4.8 on
> a monoprocessor?
>
> Thanks in advance for your suggestions and ideas!
>
> *Thibaud BACKENSTRASS* <thibaud.backenstrass at phelma.grenoble-inp.fr>
>
>
> _______________________________________________
> users mailing list
> users at rtems.org
> http://lists.rtems.org/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/users/attachments/20160527/5fb46066/attachment.html>