[ET Trac] [Einstein Toolkit] #477: Carpet commit f33caf3aba96 (hg) breaks timer output in CCTK_Terminate
Einstein Toolkit
trac-noreply at einsteintoolkit.org
Tue Jul 19 16:35:42 CDT 2011
#477: Carpet commit f33caf3aba96 (hg) breaks timer output in CCTK_Terminate
---------------------+------------------------------------------------------
Reporter: rhaas | Owner: eschnett
Type: defect | Status: review
Priority: major | Milestone:
Component: Carpet | Version:
Resolution: | Keywords:
---------------------+------------------------------------------------------
Comment (by hinder):
There is no requirement that the Cactus timers be properly nested.
However, the changes I made require that objects of the Carpet "Timer"
class are used in a nested manner. Those objects use the Cactus timers.
As far as I know, timer starting and stopping is lexically correct
throughout Carpet. i.e. every Timer.start is followed by a Timer.stop
before any timers are started without stopping them. It is possible that
something escapes out of a Timer.start .... Timer.stop sequence between
the start and stop without calling the Timer.stop code. We should figure
out what is causing this and see if we can ensure that it doesn't happen.
So in summary: it cannot be timers in the flesh that are called out of
sequence, it must be Carpet Timer objects. We should add code to that
assert statement to print the names of the two timers. The TimerNode
class has a d_name member which gives the name of the timer. We could
also add debug output to the start and stop methods of TimerNode to print
the names of timers being started and stopping. This should reveal the
problem. I don't actually have a checkout of the development version of
Carpet to test with at the moment so it's not a quick thing for me to look
at right now.
edit: this was written while Roland was typing the previous comment!
--
Ticket URL: <https://trac.einsteintoolkit.org/ticket/477#comment:6>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit
More information about the Trac
mailing list