ELOG GENETIS

Updates and Results Talks and Posters Advice Ideas Important Figures Write-Ups Outreach How-To Funding Opportunities GENETIS

GENETIS

Fields marked with * are required
Entry time:	Wed Nov 20 00:40:49 2019
Author*:
Subject*:
<p><em>Attendance for today's session: Cade, Julie, Alex M, Mitchell, Scott. </em></p> <p> </p> <p><u><strong>Tasks/to-do list:</strong></u></p> <ol> <li>Get loop to run up to the fitness function. <ul> <li>Right now we have the bash script setup to run AraSim by submitting them as parallel jobs. Today Cade added the "wait" function within the job in order to make sure the loop doesn't continue to the next step before the Arasim data is ready. Cade will give us an update on Thursday letting us know whether or not this completed or gave an error. </li> </ul> </li> <li>Fix error at fitness function.  <ul> <li>Currently, the loop stops at the fitness function calculation, and aborts (program fitnessFunction_ARA.cpp). We are unsure of why at the moment. Here are the things we are looking into.  <ul> <li>Run this outside of the loop-- ie compile the fitnessFunction_ARA.cpp (g++ -std=c++11 fitnessFunction_ARA.cpp -o fitnessFunction.exe), and run the executable(./fitnessFunction.exe). What happens? <ul> <li><span style="background-color:#FFFF00">Julie and Alex did this. It gave a segmentation fault. We need to look into this further. </span></li> <li><span style="background-color:#FFFF00">We double-checked that this .exe was properly created from compiling, and it was.</span></li> </ul> </li> <li>Double-check our AraSim output by checking: <ul> <li>Format</li> <li>File type</li> </ul> </li> <li>Is there an error past the fitness function, and it's killing the loop after this runs? <ul> <li>Is his output not properly going into the gen alg?</li> <li>Are the output files properly: <ul> <li>Filled</li> <li>Formatted</li> <li>Put in the right place</li> </ul> </li> </ul> </li> <li>Look at how it's calculating the fitness score.</li> </ul> </li> </ul> </li> </ol> <p>Since it seems like Alex and I may have found a problem at our first "test", this could be the issue. We will investigate it further. If (1) Cade's "wait" function works, and (2) we can fix this issue with the loop stopping at the fitness score calculation, ideally this means that the loop is then running. </p> <p> </p> <p>Today we also found that it is better to run on the terminal through an interactive job than a virtual desktop. This is MUCH faster -- ie we don't have to wait nearly as long as we do to get a node via a virtual desktop. This means that everyone with a PC needs to either (1) work out their issues with X11 forwarding crashing (Mitchell succeeded at this today by making sure Xquartz was set up to work via ssh to Pitzer in the settings), or (2) dual boot Linux. It seems that Mac users are not having issues, but that some of the Windows users are having their OS completely crash when XF GUI tries to forward. Users can test whether or not their X11 forwarding is working by doing the following:</p> <ol> <li> Logging onto Pitzer</li> <li>Typing "module load xfdtd"</li> <li>Then typing "xfdtd"</li> </ol> <p>This should successfully open the GUI for XF. If it crashes, users may have to do one of the things I mentioned above. Once we do succeed at having this setup, you can start an interactive job via the command: </p> <p>qsub -I -X -l walltime=1:00:00 -l nodes=1:ppn=10</p> <p>You then have to wait until the node is free. This has taken about 1/3 of the time to get than when requesting a virtual desktop. Note that, however, we may need a higher ppn (maybe something like 40). This is not totally clear yet. </p> Encoding: HTML ELCode plain Suppress Email notification Resubmit as new entry

Attachment 1:
Drop attachments here...

ELOG V3.1.5-fc6679b