Protonoc

For speed tests.

Launching on Frankenstein:

time nohup /home/paolucci/NetLogo\ 6.0.4/netlogo-headless-2.sh --model PROTON-OC.nlogo --setup-file experiments-xml/rp0.2/rp0.2-OC-speedtest.xml --table rp0.2-st.csv > rp0.2-st.out 2>&1 &


This made - on the interface - a complete run, 10k and 480 steps, in 88787.346 seconds - nearly exactly 24 hours.

Final screen:
<img width="1373" alt="Screenshot 2019-08-02 09 54 07" src="https://user-images.githubusercontent.com/1037991/62354432-2915af00-b50d-11e9-932a-77ac7257e94c.png">

Intermediate screen:
<img width="1363" alt="Screenshot 2019-08-01 09 17 10" src="https://user-images.githubusercontent.com/1037991/62273675-66f7d200-b43e-11e9-9ebd-d03d42a03c66.png">


BEGIN PROFILING DUMP
549
Sorted by Exclusive Time
550
Name Calls Incl T(ms) Excl T(ms) Excl/calls
551
FIND-ACCOMPLICES 650 10854.229 8238.593 12.675
552
SETUP-SIBLINGS 1 3164.433 2761.314 2761.314
553
PERSON-LINKS 682 1477.262 1477.262 2.166
554
POTENTIAL-FRIENDS 39017 1266.104 1129.317 0.029
555
GO 40 19828.843 847.026 21.176
556
UPDATE-META-LINKS 16 813.368 775.362 48.460
557
SOCIAL-PROXIMITY-WITH 27591 522.406 462.604 0.017
...
836

This release is invalidated by the bug addressed in e295bdac2564fc09321276dbf62b17da812d383e

This is the base of the runs that will be shown in Palermo. We have collected a few stats, and it runs with a profile of 148s on Air, and 77.4 on drake1.

I report here a few useful launchers:

drake sandman
nohup /home/paolucci/NetLogo\ 6.0.4/netlogo-headless.sh --model PROTON-OC.nlogo --setup-file experiments-xml/run-at-1000/exp-OC-baseEduEco-sandman.xml --threads 16 --table table-output-sand.csv > netlogo-sand.log 2>&1



drake profile
nohup /home/paolucci/NetLogo\ 6.0.4/netlogo-headless.sh --model PROTON-OC.nlogo --setup-file experiments-xml/OC-profile.xml --table table-output-prof.csv > netlogo.log.prof 2>&1 &


Air, profile

/Applications/NetLogo\ 6.0.4/netlogo-headless.sh --model PROTON-OC.nlogo --setup-file experiments-xml/OC-profile.xml --table table-output-profile.csv > netlogo-profile.log 2>&1 &

This release is meant to produce a first batch of results with 1000 agents.

It also contains a test for 2K agents; the command to run that test (in unix) is


nohup /opt/netlogo/6.0.4/netlogo-headless.sh --model PROTON-OC.nlogo --setup-file experiments-xml/exp-OC-test2000.xml --table table-output-2000.csv > netlogo.log.2000 2>&1


The first part should be changed to run under Windows.

This release is invalidated by bug 29256a4a13aa6f2074ac5d2218a5ccb205258800.

This release is meant to produce a first batch of results with 1000 agents.

It also contains a test for 2K agents; the command to run that test (in unix) is


nohup /opt/netlogo/6.0.4/netlogo-headless.sh --model PROTON-OC.nlogo --setup-file experiments-xml/exp-OC-test2000.xml --table table-output-2000.csv > netlogo.log.2000 2>&1


The first part should be changed to run under Windows.

We need to get some data and we need them now! So let's try this.

This should have been 1.0 but it takes four hours to run two steps.

So we need to optimize. Specifically, on labss-simul, times for the basic test are:

2000 agents
setup: 80 s
setup+ ten steps: 1718 s

that makes about 21 hours/simulation

10K agents
setup: 4812 s
setup + ten steps: 35994 s

10k agents on docker
setup: 4812
setup + ten steps: 54119 s

this means about than 1h/step. A simulation is 40 years, so 480 steps - 27 days in the worst case.