|
From: | Doug Donalson |
Subject: | Hypothetical Question |
Date: | Mon, 9 Oct 2000 17:23:14 -0700 |
I am trying to track down a fairly bizzare scheduling bug
and am trying to come up with possible senerios to account for it.
Situation: Concurrent event schedule with auto
drop. This is my asynchronous time schedule so that events for agents are
scheduled in advence and the schedule always has several thousand events already
scheduled at various times in the future.
Question: Suppose I am executing an event at time
1262.536894 and in the process a new event is generated with the exact time
value as the event presently being executed. How does the scheduler handle
this?
The problem I am chasing is that after 1260 time steps,
with at leat 100 events per time step and over 90,000 agents created and
destroyed (translated this means that this is, at least at first glance, a
unique situation) the schedule suddenly goes fubar for one time step. As
the scheduler moves through time step 1262, the time suddenly gets stuck at time
1262.253566. That is, each time an event is executed I go to the event
schedule and find the two key values for the concurrent schedule and suddenly
every executed event has the same two key values. When one of these
events schedules a new event that should be executed later in time step 1262, it
never shows up on the schedule, although the schedule returns a
non-NULL id when the event is scheduled. Also, a dump of
the event schedule at the beginning to time step 1262 shows at least 100
events scheduled between 1262.253566 and time step 1263, but at the end of the
long flurry of 1262.253566 events nothing else occurs until the time
step 1263.
I should mention that this is not just an instrumentation
error as one of the time step 1262 events that doesn't get executed causes a
fault in time step 1263.
Anyway, one sort of boundry condition that would be a rare
ocurance would be the senerio I described at the beginning of this note. I
am not implying it is happening, I am just trying to come up with some possible
hypothesis to test.
Cheers,
D4
*********************************************************************
* Dr. Doug Donalson Office: (805) 893-2962 * Ecology, Evolution, Home: (805) 961-4447 * and Marine Biology email address@hidden * UC Santa Barbara * Santa Barbara Ca. 93106 ********************************************************************* * * The most exciting phrase to hear in science, the one that * heralds new discoveries, is not "EUREKA" (I have found it) but * "That's funny ...?" * * Isaac Asimov * ********************************************************************* |
[Prev in Thread] | Current Thread | [Next in Thread] |