how valid are the test *.scn files?
Joel Sherrill
joel.sherrill at OARcorp.com
Fri Mar 31 16:28:44 UTC 2000
The screen files are generally valid. In this case, I
think the screen in 4.0 was wrong. I think this was
due to a patch to avoid printing in the middle of a
context switch user extension. I have attached the
current screen.
In general, there are a handful of factors that
diminish their accuracy.
1. Time to do output. If you are using polled IO, then
the task can spend more/less time than the test was
originally designed with. Say you are outputting at
1200 baud. The times may skew.
2. Interrupt driven IO. The tests are VERY particular
about context switch placement. This is intentional.
They set up specific scenarios. If an interrupt
from serial IO causing an unexpected context switch,
then the test IO can vary.
3. Extra tasks/semaphores, etc. in the BSP skew the
reported IDs.
4. The most common is that addresses printed vary based on
the target.
I have a patch to do IO to a memory buffer and then dump
it at known points in the tests. This will fix the IO
related issues.
--
Joel Sherrill, Ph.D. Director of Research & Development
joel at OARcorp.com On-Line Applications Research
Ask me about RTEMS: a free RTOS Huntsville AL 35805
Support Available (256) 722-9985
-------------- next part --------------
*** TEST 4 ***
TA1 - rtems_task_suspend - on Task 2
TA1 - rtems_task_suspend - on Task 3
TA1 - killing time
TA1 - rtems_task_resume - on Task 2
TA1 - rtems_task_resume - on Task 3
TA1 - rtems_task_mode - change mode to NO RTEMS_PREEMPT
TA1 - 09:15:10 12/31/1988
TA1 - 09:15:11 12/31/1988
TA1 - 09:15:12 12/31/1988
TA1 - 09:15:13 12/31/1988
TA1 - 09:15:14 12/31/1988
TA1 - 09:15:15 12/31/1988
TA1 - rtems_task_mode - change mode to RTEMS_PREEMPT
TA1 - 09:15:00 12/31/1988
TA2 - 09:15:04 12/31/1988
TA3 - 09:15:05 12/31/1988
TA1 - 09:15:06 12/31/1988
TA2 - 09:15:07 12/31/1988
TA3 - 09:15:08 12/31/1988
TA1 - 09:15:09 12/31/1988
TA2 - 09:15:16 12/31/1988
TA3 - 09:15:17 12/31/1988
TA1 - 09:15:18 12/31/1988
*** END OF TEST 4 ***
More information about the users
mailing list