channelflow.org

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Hyatt research blog, spring 2009 ======


====== Channelflow installation ======
Append the latest posts at the bottom of this page

===== Hiccups =====

:-D Having problems getting cygwin to work - John G, remember that long path that needed to be specified to get commands to work?  I can't figure out what that was supposed to be, and pressing the up arrow didn't bring it back when I first started cygwin.  I've spent almost an hour not being able to get commands to work - I'm going to keep trying, but if anyone has any idea what I need to do please let me know.  

I'm also not sure I'm adding to this blog properly - am I supposed to be using the "Edit" function, or is there a better way?
 --- //John Hyatt 2009-01-31 05:05//

8-) Yes, click on [Edit this page], then [Preview]; when satisfied, enter the edit summary, and then [Save] 
 --- //[[predrag.cvitanovic@physics.gatech.edu|Predrag Cvitanovic]] 2009-02-01 05:08//

8-o You might need to run ''export PATH=$PATH:~/channelflow-1.3.3/bin'' in your shell and perhaps 
''export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:~/channelflow-1.3.3/lib'' as well. See [[docs:INSTALL]].
Please post the exact test of the commands you are trying to run and the error messages you get. That 
will help me diagnose the trouble. //John G 2009-02-03 9:49//

When running through the tutorial, where do you download the LB.ff file and how do you access it once it's there?  Thanks - //[[jshyatt@gatech.edu|John Hyatt]] 2009-02-03 15:18// 

8-o I can't answer this exactly but I can provide some clues.
The issue is to understand how the cygwin shell presents the Windows filesystem, that is, what is the cygwin unixy 
textual representation of the Windows filesystem, which (perhaps) you are more used to thinking of as a series of nested 
folders that you click through. At the cygwin prompt, run 'pwd' (print working directory). That will tell you where 
the cywin shell starts up within the Windows filesystem. The directory (folder) names between the slashes should be 
familiar to you. If you know where you downloaded the LB.ff in terms of Windows folders, try to extrapolate from the 
folder names to a unixy textual path, so that you can copy LB.ff to the cygwin working directory. Then in cygwin you would
type something like <code> cp /Users/John_Hyatt/Desktop/LB.ff .</code> (in unix "." stands for the current directory).
I will try to find a good basic intro to unix shell usage on the web and provide a link.
 // John Gibson 2009-02-05 9:43//

OK, I'll give it a shot when I can, and let you know how it goes.  
-//[[jshyatt@gatech.edu|John Hyatt]] 2009-02-10 12:59//

:-| to John G (or whoever):  can you help me get set up on the campus Linux machines sometime?  Do they work the same as the Cygwin program on my laptop?  Where are they?  Etc.  //[[jshyatt@gatech.edu|John Hyatt]] 2009-02-10 13:04//
\\
{{gtspring2009:pc.jpg?10 }} I can open an account for you on the CNS network - students of 5th floor Howey can show you how to use it. Perhaps best to talk about this on skype?  --- //[[predrag.cvitanovic@physics.gatech.edu|Predrag Cvitanovic]] 2009-02-10 23:12//
\\
{{gtspring2009:gibson.png?10}} You probably want an account on the [[http://share-it.gatech.edu/oit/pace/user-guide|PACE cluster]]. The user interface is much like Cygwin (typing into a Unix shell) but there are a few hoops to jump through to get it set up. More documentation coming ...
//John Gibson 2009-02-11 9:50 EST//



====== Nite fishin' ======
Append the latest posts at the bottom of this page

{{:gtspring2009:pc.png|}} to John H: What [[gtspring2009:research_projects|project]] would you like to focus on (does not have to be any of the ones listed so far)?  --- //[[predrag.cvitanovic@physics.gatech.edu|Predrag Cvitanovic]] 2009-02-08 13:57//

In complex analysis you can map different-geometry systems onto one another, so you can solve a problem with an easy geometry and then map the solution onto the same problem with more complex geometry.  (A straight temperature gradient on a square heat conductor with opposite hot and cold edges is easy to solve for; if you solve it you can map the answer onto rectangles, semicircles, etc. as long as they have similar boundary conditions.)  Is it possible to do something similar with the plane couette flow?  Say, map one cell of it onto a cylinder with opposite velocity fields at the edge and at the center?  Like fluid flowing through a pipe with a stream flowing through the middle in the opposite direction - intuitively it seems like it ought to behave the same as plane couette flow, since it has one b.c. for each direction of flow, and fluid in between.  

1) Is something like that meaningful as far as "research" goes?  
2) I'm completely ignorant as far as computer knowledge I would need for this.  

PS:  I don't know how much time I will have to work on this right away - I have been doing homework in this building until 9PM 7 days a week for the past few weeks and still haven't figured out how to use the Linux environment yet.  I'm asking John Gibson to help me figure out how to use the Linux machines on campus next time he's here.  -//[[jshyatt@gatech.edu|John Hyatt]] 2009-02-10 12:47//

{{:gtspring2009:pc.png|}} Way too ambitious, methinks - best to do something narrowly within the study group goals. Finding a new periodic orbit, plotting it in states space would be already great, for example.  --- //[[predrag.cvitanovic@physics.gatech.edu|Predrag Cvitanovic]] 2009-02-12 15:46//

OK, I'll try to figure something out on Tuesday, then.  -//[[jshyatt@gatech.edu|John Hyatt]] 2009-02-23 09:54//

{{:gtspring2009:pc.png|}} Steps of the nite fishin' project: --- //[[predrag.cvitanovic@physics.gatech.edu|Predrag Cvitanovic]] 2009-03-20//
  - feed LB equilibrium data set from Channelflow into JFG's ''[[:docs:utils:findorbit|findorbit]]'', see whether it accepts it as an equilibrium
  - degrade LB equilibrium data set by dropping higer-order coefficients, see whether ''findorbit'' still find it
  - learn how to transfer data sets to Matlab or Mathematica for visualization
  - plot a movies of `turbulent' flow
  - start fishing - run guesses picked out from long-time simulations, see whether you find known equilibria, or?
  - if successful, ran a series of searches on PACE
  - blog everything as you go along

8-o I just finished running findorbit on the upper and lower branches.  It takes about 2-3 minutes for the program to run with starting guesses at the equilibria.  I don't know how to interpret the output, though - it gives a convergence.asc file, a findorbit.args file, and ubest.ff and unewt0.ff.  I couldn't find documentation on what this means on the website, but I'm guessing that unewt0.ff is the first of many steps that would be done starting farther from the equilibrium, and ubest.ff is the final equilibrium?  JohnG, could you confirm this and tell me what the easiest way to compare these with the actual solutions to make sure they are the same?  //17:26//

Update:  Have been trying to figure this out for an hour now, with no success, so I'm giving up for the time being.  I'd also like to know how to make small changes to the equilibrium .ff files, to get a feel for how long it can take the program to find an equilibrium it is "close to."  For now I'm going to try the command on a random field and see what happens.  

Also, what is the command to rename a file or directory?  //18:39//

Update:  findorbit just finished running on my first try at a random field, saying "GMRES failed to converge," and quit.  I'm going to try again for the same field, increasing the number of iterations (I think that's why it didn't work).  //20:52//

Update:  input command "findorbit -eqb -Nn 50 random1.ff" seems to have still used the default 20 number of Newton steps, same as "findorbit -eqb random1.ff."  I was trying to get it to use 50 instead of 20 since the sequence does seem to be slowly converging, but this didn't seem to work.  That's it for tonight.  //23:51//

-//2009-03-21 [[jshyatt@gatech.edu|John Hyatt]]//
channelflow.org

User Tools

Site Tools

Page Tools
