On 11/20/2014 12:08 PM, Sebastian Huber wrote:
On 20/11/14 10:22, Daniel Hellstrom wrote:
I will fix this. I missed it since it I never enabled RTEMS_PROFILING.
I did a test suite run on NGMP with profiling enabled and your patches with a local fix. So overall they don't make things worse and the median of all maximum thread dispatch disabled times drops a
bit (Std = Git master, Opt = with your patches).
How should I interpret this, is it all RTEMS test-suite you have run performing
the profiling?
Turning on/off interrupts should be faster, but the code calling the on/off
routines should also be faster. I used GRMON to extract instruction traces for
window overflow/underflow handling to benchmark.
DanielH
_______________________________________________
devel mailing list
devel@rtems.org
http://lists.rtems.org/mailman/listinfo/devel
