search

unige-geohealth / accessmod

accessmod 5 : anisotropic accessibility analysis.
GNU Lesser General Public License v3.0
42 stars 14 forks source link

Accessibility analysis: Estimated calculation time way off the real calculation time #322

Open SteeveEbener opened 4 years ago

SteeveEbener commented 4 years ago

Current Behavior

I run the accessibility analysis on the dataset posted in the following project file: https://www.dropbox.com/s/522eaabmbaor9bf/LAO_CONNECT.am5p?dl=0 using the following input parameters:

The validation window gave me the following estimated calculation time to conduct the analysis: image

I anyway decided to run the analysis and here is the time this really took (sorry for the quality of the image): image

It might be worth to look into how the estimated calculation time is being measured as the information provided might discourage some users to run the analysis while it might actually be quite quick.

SteeveEbener commented 4 years ago

I just run the same analysis but on a different dataset and got the same estimated calculation time:

image

fxi commented 4 years ago

"22 days, 18h and 7 minutes" refers to the maximum travel time that could be stored in a cell (2^15-1 minutes), not the computation time. Computation time is not predictable yet in AccessMod, except if there is a repetition of an analysis with similar data on the same computer, same RAM, same CPU.

SteeveEbener commented 4 years ago

Ho, I see. I got mislead by the message. I then would like to suggest that we combine the two associated information messages into one as follow:

"Maximum travel time set to 0 minutes. The maximum computed travel time will be of 32'767 minutes (22 days, 18h and 7 minutes)". Any cells located beyond such travel time will be coded as -1."