Bug in measuring average content

Hi,

I measure the average content of a processor with a PFM in the Experimenter. Unfortunately, there is perhaps a bug in the PFM. The replication results of every scenario is divided into two group which are very close to each other but very far from the other groups.

Example:

For each scenario, I got 20 replication. The mean content of the total 20 replications is shown as 0.5. But if I look at the single results of each replication, I found that about 10 replications have a mean content of 0.95 and the other 10 of 0.5. This look VERY strange.

Hope someone can help!

Thanks

Think you did the experiments more than once. Allways the same problems? In the same number of replications?

Exactly! But just the content PMF of the processors, not of the queues.

It´s hard to reproduce what you´ve done without a example. I´ve built a small example model and get results that seems to are ok.

here I got a simple example. Just take a look on the PMF of the average content of a processor. Thanks!

This problem is being caused by a bug in the warmup time statistics reset. The average content is not being precisely calculated when there is a warmup time, particularly when the warmup time is a large portion of the overall simulation run time.

When there is no warmup time, this problem does not occur, which is probably why Sebastian was unable to duplicate the problem.

We will fix this bug for the next release (5.1.0), which should be available sometime this week.