Add information summarizing computational load about mesatesthub HOT 9 CLOSED

Running a little test in the database console, I get 1.13e5 cpu hours for cannon over its lifetime, so it looks like this should work and this won't require much retooling to get on the testhub.

from mesatesthub.

Cannon usage over the last 6 months.

from mesatesthub.

Are we just interested in the inner product of threads and runtime? That should be quite easy to throw on the computer view. It won't capture compile time, but I assume that's a perturbation. How about core hours for the following time spans:

• last 24 hours
• last 365 days
• lifetime

from mesatesthub.

I think I have a working implementation. It is expensive and slow for very prolific submitters like helios and cannon, since it must load all test instances from the database and computer the inner product in series. Obviously, this won't scale well, so in the future, I'd like to keep a running total handy directly in the database updated once a day or with every instance submission.

For now, here's an example for lux. See the data on the right side of the view. Can someone let me know if this is what we're looking for?

from mesatesthub.

this will be useful for me later this year. two clarifications please. does "Last year" mean the past 365 days of today? what is the start date for "Since Inception" - since github day or when the machine first submitted a result?

from mesatesthub.

Last year means since this time on this calendar date a year ago, and since inception means since this computer stated testing git commits. This does not include svn-era data. That database is archived on a computer in Eau Claire, but not particularly easy to access.

from mesatesthub.

thanks. might be useful to add the date of "since inception", or the elapsed number of years "since inception". one could then get the average load per year and compare that to the "last year" to get a sense of the distribution.

from mesatesthub.

Sounds good. I'm first finding a more database-dependent way to compute these values... This first attempt is crushing the server.

from mesatesthub.

This took longer than expected, but every test instance with a nonzero number of threads and a nonzero runtime now knows its cpu usage, and the computer summary pages show the indicated data, and the inception date for each computer is clearly indicated.

from mesatesthub.