We need a better description of the model run times

[gregmacfarlane]

Reviewer 1 says:

I was quite surprised that the demand components of the trip based model took 10 hours, that seems very long. However, the flow chart shows that an assignment is part of each iteration which is somewhat more reasonable, but still quite long. But, this is not an even comparison with the ActivitySim that does not include any network initialization, building, or skimming steps. A better comparison could be a component by component or a comparison of the number of computations.

I think I agree with this criticism. Can we get a breakdown of the WFRC model run time by:

• How long does skimming / assignment take?
• How long does one loop of trip generation / distribution / and mode choice take?

This is of course made more difficult by the fact that the WFRC model does mode choice AFTER assignment has converged. But can we get a log and build a schedule?

Would be nice to get an apples-to-apples (one run of ActivitySim, one run of the demand components of the WFRC model)

[gregmacfarlane]

ActivitySim log is here: https://byu.app.box.com/file/1292154340611
WFRC log is here: https://byu.app.box.com/file/1405569820101

[gregmacfarlane]

Perhaps we can build a figure that has better details of the execution at each step, removing feedback loops.

[gregmacfarlane]

Actually, let's use the activitySim log at C:\projects\wfrc_asim_scenario\activitysim\output\base_2019_mp_3 on Beast