Updates and Results Talks and Posters Advice Ideas Important Figures Write-Ups Outreach How-To Funding Opportunities GENETIS
  GENETIS  ELOG logo
 List  |  New  |  Edit  |  Delete  |  Reply  |  Duplicate  |  Find  |  Config  |  Help 
This is a draft message, edit and submit it to make it permanent  
Message ID: 36     Entry time: Tue Feb 25 17:10:36 2020
Author: Alex Machtay & Alex Patton 
Subject: Working Meeting 2/25/20 

Today's Tasks

Mitchell, Evelyn, Eliot, Ryan: Work through the bash project; Mitchell is done, Ryan and Evelyn are close to being finished.

Leo: Worked on next presentation (not sure when he'll present--either Friday or next Tuesday).

Alex M: Continued commenting code and helped Alex P. with changes in Part_B and Roulette_algorithm .

Alex P: Fixed errors in loop, tested with an interactive job (using CPU). Errors look to be fixed up through part B. Arasim should be launching and using the $tmpdir, though we need to get further in the testing to see for sure.

Made a copy of the loop that functions without GPUs so that if GPU is unavailable some progress can be made even if it is slow.

While waiting for a GPU Alex P ran on CPUs, at half scale factor one indiviudal in XF took 32 minutes (although it was predicted to be over 2 hours). 

To Do List:

We want to keep testing the loop. Right now, we are using a cpu running at half scale. This gives estimated times for each antenna of up to 2 hours. GPUs are suddenly hard to come by, so if we are relegated to using CPUs we'll probably need to scale down considerably.

Evelyn, Ryan, and Mitchell are probably going to be ready for the python project soon. If someone is around on Friday to help them they should be able to finish the bash project.

One idea is to run all the inital xmacros in the interactive job and then submit a GPU job that runs all the xfsolver command. With the hope that not having the GPU in the interactive job will make it easier to get the interactive job and to get the GPU for a shorter time because it would only be needed for a smaller segment. This would especially help with running larger amounts of individuals.

 

ELOG V3.1.5-fc6679b