Bugs in new versions of Athena and Artemis
Hello, I just installed the new update so now I have Athena 0.8.026 and Artemis 0.7.001. First I would like to say that I like the new features, particularly the possiblity to run Atoms and FEFF within Artemis and the fact that all files are included in one zip-file. However, I noticed some major bugs on my Windows 2000 machine (although I do not want to exlude the possibility of wrong handling). I attached some example files as a zip archive. Athena: - I usually read in a rather large number of chi(k) data-files. It then happens often that a few of the created groups will not be displayed: when I mark/select such a group and press a purple/red button it might for example plot correctly for E, R and q; but in k-space it will simply give a message to the Echo Area ( .. plotting in k-space .. done! ) but the resulting graph will be blank. This behaviour generally occurs when the number of data-files is around 150 but I also observed it with less files. I am sure that the files to import are corrupt as they behave well when I read them in separately. - When I close the graph window the main athena program will crash. I remember that this behaviour has been discussed earlier. - While loading (and saving) of projects with many data groups is reasonably fast, closing such a project requires an extremely long time. Is that a normal behaviour? Artemis: The behaviour of Artemis on my machine makes its use almost impossible: - Data file can not be read in: When I read in a datafile (Data | Change data file) Athena will correcly show up a dialog and display the data in k-space in the graph window. Any subsequent click on k, R or q-buttons will give the error message: "You have not yet loaded data ...". I am not be able to save the datafile, instead an error message will occur "You cannot write the artemis description file". I tried it with my own data (exp005_absW.chi) and with an example file (cu010k.chi). If I first set up a FEFF calculation and afterwards import a data file I am able to plot the data file in k-space but in R-space it will only show a horizontal line. When fitting always a warning message will appear: "There are errors in your math expressions ...". "Co Cubic 3.apj" is an example of such a file. As you can see, the folder chi_data in the zip-file contains no data. - Having multiple FEFF-Calculations in one file is not possible: 'Co Hexa Cubic.apj' is a project where I imported two different atoms calculations (first hexagonal cobalt, second cubic cobalt) and ran FEFF. But only the second calculation can be displayed. And when looking at the project file there exists only one FEFF-calculation despite two trees in the Artemis program. It seems as if the first calculation has been overwritten. - Previous FEFF paths are not removed from Zip file 'Co Cubic 1.apj' is a project where I first set the cluster size to 8 and then ran Atoms and FEFF and imported all paths. I then set the cluster size to 4 and did the procedure again. The second time the same number of paths will be imported as in the first case although there should only exist a fraction of them. This problem can only be solved by deleting first all paths in the zip-file manually (which I do using a zip-program) and then run FEFF again. - Making 'Sum of paths (no fit)' produces no output (when no data file is loaded). In 'Co Cubic 2.apj' I set up an Atoms page and ran the FEFF calculation. When choosing 'Sum of paths (no fit)' | all paths from the menu it will indicate a correct calculation but when trying to plot the fit it will fail. I wonder if the described behaviour depends on Windows or on my particular computer. I tried a completely new installation which did not improve the situation. After complaining so much I would still like to propose a new feature: after processing data with Athena I export the chi(R)/chi(q) data to use it in a data analysis program with enhanced graphing capabilities. I would really appreciate if the used parameters would be written at the beginning of the exported datafile as it is quite difficult to keep track of all these options. I hope my remarks are helpful. Regards, Gerrit
On Thursday 11 March 2004 08:43 am, Schmithals wrote:
Athena: ]
I'll address the comments on Artemis in a subsequent email.
- I usually read in a rather large number of chi(k) data-files. It then happens often that a few of the created groups will not be displayed: when I mark/select such a group and press a purple/red button it might for example plot correctly for E, R and q; but in k-space it will simply give a message to the Echo Area ( .. plotting in k-space .. done! ) but the resulting graph will be blank. This behaviour generally occurs when the number of data-files is around 150 but I also observed it with less files. I am sure that the files to import are corrupt as they behave well when I read them in separately.
This has been discussed in the past on the mailing list. As you all know, Athena is just a front-end for Ifeffit. Ifeffit is written in Fortran and therefore has hardwired limitations on the amount of memory it can access. Athena makes a rather anemic attempt to estimate how much memory will be needed and a similarly anemic attempt to warn you as you approach the memory limit. This is better than nothing, but is flawed. I suspect that 150 files is too many. That is, 150 times the memory needed to store the energy, k, and R arrays is more than Ifeffit allows. Once you exceed this memory limit, weird stuff will happen. The best suggestion I can make is to split your data up into groups of 40 or 50 files and save projects files with only that number. If you need to merge 150 files, the best suggestion is to merge them in groups of, say, 20 then merge the merged groups.
- When I close the graph window the main athena program will crash. I remember that this behaviour has been discussed earlier.
Read the FAQ: http://cars9.uchicago.edu/cgi-bin/ifeffit/faqwiz?req=all#4.3
- While loading (and saving) of projects with many data groups is reasonably fast, closing such a project requires an extremely long time. Is that a normal behaviour?
It is normal behavior. And this behavior has been bugging me more and more lately, so it will prbably be improved in the near future. B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 405 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
Gerrit I am looking into each of your points about artemis. I have not seen any of these behaviors on my linux machine, so I booted up the Windows 2K virtual machine and am checking them out. Just to keep my thoughts organized, I am going to address each of your points in separate emails as I get to them.
- Having multiple FEFF-Calculations in one file is not possible: 'Co Hexa Cubic.apj' is a project where I imported two different atoms calculations (first hexagonal cobalt, second cubic cobalt) and ran FEFF. But only the second calculation can be displayed. And when looking at the project file there exists only one FEFF-calculation despite two trees in the Artemis program. It seems as if the first calculation has been overwritten.
I concur that the project file you sent is flawed. There is certainly only one feff calculation in the file. However, I cannot reproduce this problem on my computer. When I import two atoms.inp file and run atoms then feff in any order, I end up with two usable feff calculation. When I save the project and reopen the project, I still have two usable feff calculations. Apparently something went wrong as you were doing this, but I don't yet have enough information to track down the problem. A few questions: 1. Did you do anything to the folders or files under C:\Program Files\Ifeffit\horae\stash\artemis.project.0\ while Artemis was still running? 2. On the taskbar is an item that says "artemis" and that has the same icon as the desktop icon. Can you click it open and see if any kind of error messages are written there while you reproduce this problem? 3. When you reproduce this problem, keep careful track of which buttons you push and in which order. That may be useful information for me. On to the next problem... B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 405 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
On Thursday 11 March 2004 08:43 am, Schmithals wrote:
- Making 'Sum of paths (no fit)' produces no output (when no data file is loaded). In 'Co Cubic 2.apj' I set up an Atoms page and ran the FEFF calculation. When choosing 'Sum of paths (no fit)' | all paths from the menu it will indicate a correct calculation but when trying to plot the fit it will fail.
This project file is triggering several repititions of the same bug. The bug (my fault!) is triggered by not having data in the project. I believe that you can avoid this issue simply by importing a data file. It doesn't even have to be a relevant data file if all you want to do is look at the sum of paths. I acknowledge that one can reasonably want to do a sum of paths without importing data. It is my intent that such a thing should be possible and I will work to make it so in the next release.
- Data file can not be read in: When I read in a datafile (Data | Change data file) Athena will correcly show up a dialog and display the data in k-space in the graph window. Any subsequent click on k, R or q-buttons will give the error message: "You have not yet loaded data ...". I am not be able to save the datafile, instead an error message will occur "You cannot write the artemis description file". I tried it with my own data (exp005_absW.chi) and with an example file (cu010k.chi).
Again, I cannot reproduce this problem. This simply does not happen for me. Are there any interesting messages in the command window?
If I first set up a FEFF calculation and afterwards import a data file I am able to plot the data file in k-space but in R-space it will only show a horizontal line. When fitting always a warning message will appear: "There are errors in your math expressions ...". "Co Cubic 3.apj" is an example of such a file. As you can see, the folder chi_data in the zip-file contains no data.
I can reproduce this starting with your project file, but I cannot do so starting from scratch, saving the project, reopening the project, then reading some data in. I did notice that, when staring with your project file, the kmax for the Fourier transform to 0, then the data will plot in k but will be a horizontal line at 0 in R. That seems like correct behavior to me. Did you check your k-range values? Setting kmax to something sensible makes for a sensible FT. As for the fitting message -- I should check to see that kmax > kmin and Rmax > Rmin. I currently do not make that check. B The basic flowchart for an Artemis 0.7 project is this: (1) import data, (2) import or create atoms data, (3) run atoms, (4) check the feff.inp data and run feff, (5) set up your fitting model, (6) hit the big green button. Certainly you should be able to deviate from that script, but if you run into some problem with the code, it may be prudent to follow it. B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 405 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
On Thursday 11 March 2004 08:43 am, Schmithals wrote:
After complaining so much I would still like to propose a new feature: after processing data with Athena I export the chi(R)/chi(q) data to use it in a data analysis program with enhanced graphing capabilities. I would really appreciate if the used parameters would be written at the beginning of the exported datafile as it is quite difficult to keep track of all these options.
Do you mean just the data parameters (krange, Rrange, and so on) or do you mean some kind of description of the entire fit? You can edit the titles on the data page to put your own information in the header written to the data chi(x) files. But I see your point and will include that in the next release. B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 405 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
On Thursday 11 March 2004 08:43 am, Schmithals wrote:
- Previous FEFF paths are not removed from Zip file 'Co Cubic 1.apj' is a project where I first set the cluster size to 8 and then ran Atoms and FEFF and imported all paths. I then set the cluster size to 4 and did the procedure again. The second time the same number of paths will be imported as in the first case although there should only exist a fraction of them. This problem can only be solved by deleting first all paths in the zip-file manually (which I do using a zip-program) and then run FEFF again.
Running feff repeatedly is indeed a problem in the current version. In fact, there are issues far beyond the one you bring up here. It is going to take me quite a while to fix all of the problems with running feff repeatedly for the same feff calculation in an artemis project. There are some seriously subtle issues involved. The best advice I have for now either: (1) Completely delete the feff calculation from the project using the discard option in the Feff menu and then start over again or (2) Set the cluster size bigger than you anticipate needing and just don't worry about all the extra paths being carried around in the project file. B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 405 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
participants (2)
-
Bruce Ravel -
Schmithals