Issues from impromptu VisIt user's meeting at JOWOG-43

I've put all these issues here as a placeholder for now. After we've had a chance to discuss them, we can file individual tickets as appropriate. Font issues text overlapping text running off screen text not scaling well as image size is varied text not saving well when saving images Python Code Examples A lot of our examples apparently using older python (1.7) Would like examples updated to use newest python I explained 3.0 was a challenge for us due to some major changes Update to CLI manual? Session File issues major pain moving session files between machines (if with using restore with different sources) AWE users have given up on session files entirely, they do everything manually Recommend teleconf with Paul Selby / Matt Wheeler to collect from the usage issues Recommend following up with Josef Crepeau to get more details on session file issues In client/server session file is saved to client and would like to save to server DOD HPC Web Would like to integrate VisIt better there / support it better I don't really know what this is Additional Issues / Details Use the extra CNTA funds we have to improve the user interface / invest in useability. Problem where the fonts are specified as a percentage of the screen size, so when you switch to a different monitor they can change in size and sometimes run off the window. Wants to save his session file in the same place that his data is, not on his local machine (I guess hes running in client-server mode). Session files are difficult to restore correctly (cant be shared). Two users were in the same room with a seemingly identical connection and couldnt share session files successfully. I could have sworn I had trouble using the same session file I saved in the Gnome Desktop virtual application on the HPC portal in the SRD environment. Since it takes a long time for my images to render, Id like to be able to restore a session with sources and not have it automatically re-render all the images. I was trying to make a tiled movie, and it didnt seem to matter much whether I used 1 processor or 16 processorsapparently, based on the timing of the messages I was getting, it was taking about 80% of the run time just loading in our tecplotlike .plt files. I couldnt ever keep the connection open long enough to make the whole movie (I estimated 60 hours. Each file was about 9GB.), so I had to settle for simple movies of each tile. Each of those took about 7 hours to put ~60 frames into a .wmv movie with a 0.5 second dwell time for each frame. Perhaps I need to switch to 2000 processors. :-)

-----------------------REDMINE MIGRATION----------------------- This ticket was migrated from Redmine. As such, not all information was able to be captured in the transition. Below is a complete record of the original redmine ticket.

Ticket number: 2131 Status: Developer Review Project: VisIt Tracker: Bug Priority: Normal Subject: Issues from impromptu VisIt user's meeting at JOWOG-43 Assigned to: - Category: - Target version: - Author: Mark Miller Start: 02/06/2015 Due date: % Done: 0% Estimated time: Created: 02/06/2015 12:43 pm Updated: 03/10/2015 04:38 pm Likelihood: 3 - Occasional Severity: 2 - Minor Irritation Found in version: 2.8.2 Impact: Expected Use: OS: All Support Group: Any Description: I've put all these issues here as a placeholder for now. After we've had a chance to discuss them, we can file individual tickets as appropriate. Font issues text overlapping text running off screen text not scaling well as image size is varied text not saving well when saving images Python Code Examples A lot of our examples apparently using older python (1.7) Would like examples updated to use newest python I explained 3.0 was a challenge for us due to some major changes Update to CLI manual? Session File issues major pain moving session files between machines (if with using restore with different sources) AWE users have given up on session files entirely, they do everything manually Recommend teleconf with Paul Selby / Matt Wheeler to collect from the usage issues Recommend following up with Josef Crepeau to get more details on session file issues In client/server session file is saved to client and would like to save to server DOD HPC Web Would like to integrate VisIt better there / support it better I don't really know what this is Additional Issues / Details Use the extra CNTA funds we have to improve the user interface / invest in useability. Problem where the fonts are specified as a percentage of the screen size, so when you switch to a different monitor they can change in size and sometimes run off the window. Wants to save his session file in the same place that his data is, not on his local machine (I guess hes running in client-server mode). Session files are difficult to restore correctly (cant be shared). Two users were in the same room with a seemingly identical connection and couldnt share session files successfully. I could have sworn I had trouble using the same session file I saved in the Gnome Desktop virtual application on the HPC portal in the SRD environment. Since it takes a long time for my images to render, Id like to be able to restore a session with sources and not have it automatically re-render all the images. I was trying to make a tiled movie, and it didnt seem to matter much whether I used 1 processor or 16 processorsapparently, based on the timing of the messages I was getting, it was taking about 80% of the run time just loading in our tecplotlike .plt files. I couldnt ever keep the connection open long enough to make the whole movie (I estimated 60 hours. Each file was about 9GB.), so I had to settle for simple movies of each tile. Each of those took about 7 hours to put ~60 frames into a .wmv movie with a 0.5 second dwell time for each frame. Perhaps I need to switch to 2000 processors. :-)

Comments: Here are some more, not mutually exclusive notes (great recall, Mark!):1. Jeff McAninch said we could use the extra CNTA funds we have to improve the user interface / invest in useability.If that is through Dave Pugmire, great. OTOH, if you are thinking in terms of sending funding direct to LLNL, let us know what we can do to help.2. Jeff McAninch said there is a problem where the fonts are specified as a percentage of the screen size, so when you switch to a different monitor they can change in size and sometimes run off the window.Just to be sure, you're talking about text items in the main visualization (viewer) window and not fonts associated with GUI menus, buttons and dialogs, right? Is the whole notion of having vis. fonts defined such that they scale with screen size an issue? I mean, what if we just went to fonts like you see in a word processor specified by point size? One problem we see here all the time is something like the Mac Retina display. We're using VTK to handle our vis. fonts and it doesn't have any notion of 'point size' for fonts. Furthermore, it doesn't use any of the platform's underlying font management features to render them. It uses GL directly to render them. I mention this because fixing some of these issues will most likely involve work on VTK as well. We (the VisIt team) should at least identify what changes are needed though (both in VisIt and VTK).3. Joe Crepeau said that he wants to save his session file in the same place that his data is, not on his local machine (I guess hes running in clientserver mode).Yes, got that in my orig. list. Thanks.4. Everyone agreed that session files are difficult to restore correctly (cant be shared). Jeff McAninch recounted when he and Joe Crepeau were in the same room with a seemingly identical connection and couldnt share session files successfully.Based on what I was hearing, I can't agree more. This is something we should be able to fix. I kinda felt it was your highest priority. Does anyone disagree?As for me, I have a few notes of my own:1. I could have sworn I had trouble using the same session file I saved in the Gnome Desktop virtual application on the HPC portal in the SRD environment on us.navo.hpc.mil , but Im not sure that was my main problem then.Who's a point of contact for that system? I have no idea what we're talking about here and it sounds like a potentially logistically more complicated enviornment than a standard HPC LCF.2. Since it takes a long time for my images to render, Id like to be able to restore a session with sources and not have it automatically rerender all the images.Ok, we can easily fix this by adding a switch to the restore session dialog to NOT immediately render. OTOH, you can also change this behavior by just not saving a session when the plots are rendererd. First, go to the viewer window and clear all plots. Then save your session. That way, when you restore it, it won't actually draw your plots, only load them.3. I was trying to make a tiled movie, and it didnt seem to matter much whether I used 1 processor or 16 processorsapparently, based on the timing of the messages I was getting, it was taking about 80% of the run time just loading in our tecplotlike .plt files. I couldnt ever keep the connection open long enough to make the whole movie (I estimated 60 hours. Each file was about 9GB.), so I had to settle for simple movies of each tile. Each of those took about 7 hours to put ~60 frames into a .wmv movie with a 0.5 second dwell time for each frame. Perhaps I need to switch to 2000 processors. :)VisIt's ability to execute in parallel is somewhat dependent on how the data producer creates its output files. If the input tecplot file contains only a SINGLE block, then VisIt will only ever be able to use 1 processor to process it. If the input tecplot file contains say 5 blocks, then VisIt can usefully use at most 5 processors to process it. In general, VisIt does not redecompose input data to enable it to use a different number of processors than the original data is "blocked" into (this isn't entirely true, for structured grids like images or netcdf, VisIt can in fact redecompose those on the fly - but its not common). We see this often with Engineering codes that do some kind of a merge operation after writing a file per processor. The do the merge because some of their workflows can handle only a single, whole, coherent mesh object (e.g. they can't handle it in pieces), but once they do that, they subvert VisIt's ability to process that data in parallel.So, do you know how many 'blocks' that tecplot file had? I think the keyword "Zone" in tecplot defines a wholly independent mesh block in a file. Count how many of those keywords are in your file.Yes, and, VisIt doesn't bother to tell you when you have allocated more processors than it can possibly use on a particular dataset. Perhaps it should at least present a warning message to that effect?Some of these issues are fixable with a small amount of work and we can try to schedule them for an upcoming release.Some of these issues require more effort and/or other projects (like VTK). AFter I've had a chance to discuss these with the team, then I can give you clearer picture.In the meantime, one thing that would be good for your group to do, after we think we have a complete list of issues, is to order them in priority order.I have gone ahead and captured these details in our bug tracker, http://visitbugs.ornl.gov/issues/2131. We will discuss them at our next weekly meeting next tuesday. Very likely, we will submit additional issue tickets from there for specific work we can easily schedule. For other parts, we will get back in touch with you regrading priorities, etc.Thanks for the great notes! I think we did most of our work on VisIt 2.7.1 and 2.7.3, but recently its been upgraded to 2.8.1 and 2.8.2 (and I think Ive still had trouble with session files). Notewise responses:2. I havent seen the font scaling issue, but my understanding from the user meeting was that it was the fonts in the visualization windows. I think having the font size be relative to the size of the window (instead of screen) in the same way the plots scale makes sense and shouldnt be hard, and I dont see why the text would sometimes extends off the window if this were happening properly. Basically, you want a particular visualization (plot and text in visualization window) to look the same no matter where you are running VisIt, because we dont have reliable connections to the HPCs and often need to change where we are running VisIt to get the job all the way done.a. Incidentally, I have a related usability complaint: it takes me a long time to get the text and the plot where I want them to be in the visualization window, partly because when I resize the window the plot doesnt resize. I guess I expect the text to not resize and the plot to always scale up to take up the same fraction of the window as I resize the window, and that that would be the easiest way to change the size and position of the text relative to the plot (now, when I resize the window, I need to zoom in on the plot and repan to get it where I want, which is pretty slow for my lousy connection to the machine). I realize just changing the text so it is always the same size relative to the window is the opposite of what I want, but it is important to not have the problem Jeff mentioned as you switch machines. Perhaps putting some accessibly checkboxes in the toolbar on the visualization window that allow you to lock the size of the text relative to the window width and to lock the size of the plot relative to the window width would allow me to more easily resize things relative to one another. I also havent figured out how to specify the size of a window (in rows and columns of pixels) rather than just resize GUIwise.4. Agree that ensuring portability of session files sounds like the top priority.5. (1) Im not sure who the right POC for the DoD HPCs is, but Rick Angelini installed the new versions of VisIt and spent a lot of time trying to help me make a tiled movie. Angelini, Richard C (Rick) CIV USARMY ARL (US) richard.c.angelini.civ@mail.mil . We have 4 ways to run VisIt: inside a virtual Gnome Desktop, inside an SRD window, as a virtual application on the DoD HPC web portal, and in clientserver mode from our local machines. (Rick was aghast that we would try to run from inside of the virtual Gnome Desktop, incidentally.) Since Im having trouble even connecting to VisIt any of these ways on the SIPRNet, Im guessing the session issues are even worse there.6. (2) Thanks for the workaround on saving a session with the plots cleared!7. (3) Ill have to ask the developerits a binary file. More insight into how many processors to use would be good, in general. A popup notice about how many processors you could use on a dataset would be nice. I was wondering the other day how to change the number of processors in the compute engine, or to launch another one and switch to it, from inside the GUI. (I didnt want to just launch another process for my moviemaking, because I didnt want to get charged for 2 compute nodes and didnt know whether VisIt would only allocate processors from the current compute node.) Perhaps it would help to know which processors, which nodes, and how much memory VisIt was using, so I could make sure Im not wasting processing power.

visit-dav / visit

Issues from impromptu VisIt user's meeting at JOWOG-43 #2137