[visit-users] Launching Visit on Syrah

Gunther Weber ghweber at lbl.gov
Fri Jul 31 20:25:13 EDT 2015


Hi,

I investigated this a bit more here at LBL talking to the BoxLib developers. There are different BoxLib file types, differentiated by the first line in “Header”. Most of the BoxLib files that I encountered are of type “HyperCLaw" and that seems to be the type that VisIt expects. Andy Nonaka at CCSE has a copy of one of the files Mark Converse sent him. This is a file of type “CartGrid”, which seems to have additional information in it. VisIt completely ignores the first line of the file with the type. I assume when we were developing the reader in the first place, we did not have any access to “CartGrid” files so it did not make any difference.

However, if we want to read other BoxLib file types correctly, we need to read the first line and check the file type. The main difference between “HyperCLaw” and “CartGrid” files seems to be that they contain a “vfEps” (1e-08 in your file) per level between dx (the line "0.05 0.05 0.05” in your file) and coordSys (the line “0” following the “1e-8” in your file). Removing this line (or these lines if there is more than one level), should allow VisIt to read the file until the reader is modified. 

Gunther

On Jul 30, 2015, at 12:50 PM, Converse, Mark C. <converse3 at llnl.gov> wrote:
> 
> Mark,
>  
> Actually it is not working.  I was on travel last week and was vpning in, using my mac laptop.  It was working on that.  But I just got back today, and tried it on my desktop and it doesn’t work.  I think  I am now having the same problems you were having (even after removing that line with the 1e-08).  I tried it on my laptop locally (so here in the office) and it also doesn’t work.  I even took it home to try while vpn from a remote location and it doesn’t work.  I have no idea why it was working last week and doesn’t work this week.  Any thoughts?
>  
> Thanks,
> Mark
>  
> From: Converse, Mark C. [mailto:converse3 at llnl.gov] 
> Sent: Friday, July 24, 2015 7:24 AM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
>  
> Thanks for all the help :)
>  
> Mark
> From: Miller, Mark C. [miller86 at llnl.gov]
> Sent: Thursday, July 23, 2015 2:25 PM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
> 
> Oh, good news. At least something I helped with worked this week ;)
>  
> Mark
>  
>  
> From: <Converse>, "Mark C." <converse3 at llnl.gov>
> Reply-To: VisIt software users community <visit-users at elist.ornl.gov>
> Date: Thursday, July 23, 2015 2:04 PM
> To: VisIt software users community <visit-users at elist.ornl.gov>
> Subject: Re: [visit-users] Launching Visit on Syrah
>  
> Mark,
>  
> I removed that line as well, and I can open in Visit, then do a psuedocolor plot and draw a plot of that variable.  So at first blush it seems to be working.
>  
> Mark
>  
>  
> From: Miller, Mark C. [miller86 at llnl.gov]
> Sent: Thursday, July 23, 2015 11:21 AM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
> 
> Hi Mark,
>  
> I have debugged a bit more. IMHO, the problem is definitely with the Header file. The relevant section is below. . .
>  
> -2 -2 -2
> 2 2 2
>  
> ((0,0,0) (79,79,79) (0,0,0))
> 100
> 0.05 0.05 0.05
> 1e-08
> 0
> 0
> 0 75 5.57623875809867e-06
> 100
>  
> Note the '1e-08' line.
>  
> VisIt's current plugin is expecting an integer there for the 'coordSys' data member of the Boxlib3D plugin. It reads as far as the '1' and sets coordSys int to 1. Then it attempts to read what it thinks is another 'integer' which I think is indicating if ghost zones are used. It tries to read the 'e-08' part via >> operator and that operation fails (i assume) or somehow the 'integer' value that gets set is '3'.
>  
> If I take out the 1e-08 line, neither boxlib nor VisIt crash anymore. However, I still don't get to a state where I can plot anything. The mdserver belives there are 75 blocks and a bunch of variables but VisIt's GUI menus are non-responsive.
>  
> I think somehow we're dealing with different file format versions.
>  
> Mark
>  
>  
> From: <Converse>, "Mark C." <converse3 at llnl.gov>
> Reply-To: VisIt software users community <visit-users at elist.ornl.gov>
> Date: Thursday, July 23, 2015 9:47 AM
> To: VisIt software users community <visit-users at elist.ornl.gov>
> Subject: Re: [visit-users] Launching Visit on Syrah
>  
> "The problem is VisIt thinks the mesh has zero blocks."
>  
> Mark, do you think it could be the boxlib plugin?  Should we be focusing there?
>  
> Thanks,
> Mark
>  
>  
>  
> From: Miller, Mark C. [miller86 at llnl.gov]
> Sent: Tuesday, July 21, 2015 11:58 AM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
> 
> FYI, I have found and fixed a tiny bug in VisIt that allowed a mesh with zero domains to cause a malloc for negative size which is why VisIt was crashing with this file. However, that doesn't fix VisIt's ability to display this file. It only means that VisIt no longer cores trying to read the file.
>  
> When Visit reads the file, it finds stuff like this. . .
>  
> All Times are Accurate
> Times:
>         5.57624e-06
> All Cycles are Accurate
> Cycles:
>         100
> Meshes:
>         Name = Mesh
>         Number of blocks = 0
>         Block origin = 1
>         Cell origin = 0 (origin within one block of the cells).
>         Node origin = 0 (origin within one block of the nodes).
>         Group origin = 1
>         Title for domain hierarchy is patches
>         Title for individual piece in domain hierarchy is patch
>         The name scheme is: @level1,patch%d at n+1:
>         Number of groups = 1
>         Group ids are: 0: 0--1,
>         Title for group hierarchy is levels
>         Title for individual piece in group hierarchy is level
>         Mesh type is AMR Mesh.
>         Spatial Dimension = 3
>         The spatial extents are not set.
>         Topological Dimension = 3
>         Block names:
>         Disjoint elements No
>         Contains ghost zones Maybe
>         Contains original cells No
>         Contains original nodes No
>         Units =  x: "", y: "", z: "".
>         Labels =  x: "X-Axis", y: "Y-Axis", z: "Z-Axis".
>         Mesh coord type is XY
>         Mesh is primarily cell-based
>         Unit cell vector #0 is 1 0 0
>         Unit cell vector #1 is 0 1 0
> .
> .
> .
> Scalars:
>         Name = "density"
>         Mesh is = "Mesh"
>         Centering = node centered.
>         The extents are not set.
>         This variable does not contain enumerated values.
>  
>         Name = "xmom"
>         Mesh is = "Mesh"
>         Centering = node centered.
>         The extents are not set.
>         This variable does not contain enumerated values.
>  
>         Name = "ymom"
>         Mesh is = "Mesh"
>         Centering = node centered.
>         The extents are not set.
>         This variable does not contain enumerated values.
>  
>         Name = "zmom"
>         Mesh is = "Mesh"
>         Centering = node centered.
>         The extents are not set.
>         This variable does not contain enumerated values.
>  
>         Name = "eden"
>         Mesh is = "Mesh"
>         Centering = node centered.
>         The extents are not set.
>         This variable does not contain enumerated values.
>  
>  
> The problem is VisIt thinks the mesh has zero blocks.
>  
> Mark
>  
>  
>  
> From: Gunther Weber <ghweber at lbl.gov>
> Reply-To: VisIt software users community <visit-users at elist.ornl.gov>
> Date: Tuesday, July 21, 2015 11:25 AM
> To: VisIt software users community <visit-users at elist.ornl.gov>
> Subject: Re: [visit-users] Launching Visit on Syrah
>  
> Hi,
>  
> 1) The fact that AmrVis can read the file is promising and indicates that the file is ok.
>  
> 2) However, are you writing ghost zones into the file? The error message "Would like to have issued warning "Reader does not currently support ghostzones.", but no callback was registered.” seems to indicate that the file has ghost zones in it, while the reader currently does not support that, as indicated by its notes:
>  
> //   -> Ghost zones are currently not supported, and a warning is issued
> //      if a dataset is found to have ghost zones. This is due to lack
> //      of sample datasets, and should be removed once appropriate data
> //      is avaliable.
>  
> 3) BoxLib itself does not seem to have a release number itself. CCSE’s Marc Day maintains tar balls that are something like a release for a collaborating projects and a while back I migrated VisIt to use one of these TAR balls at hxxps://ccse.lbl.gov/Software/tarfiles/. VisIt uses 1.0.7, while the most recent release seems to be 1.3.0. It’s always a bit of a hassle to update, so I haven’t done it in a while. The BoxLib file format tends to be very stable, but I can check if there were any recent changes.
>  
> I would check the ghost zone issue first and then take things from there.
>  
> Gunther
>  
>  
> On Jul 20, 2015, at 7:44 PM, Converse, Mark C. <converse3 at llnl.gov> wrote:
> Mark,
> I'm not sure about the patches, but the AMR algorithm, can allow multiple mesh refinement levels.  This simulation only allows one though.
> I have a contact I'm going to ask about the Boxlib in Visit.
> Mark
> From: Miller, Mark C. [miller86 at llnl.gov]
> Sent: Monday, July 20, 2015 5:39 PM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
> Is there anything funky about the number of patches and/or levels here.
> The VisIt plugin seems to think there is one level with zero patches.
> From: <Converse>, "Mark C." <converse3 at llnl.gov>
> Reply-To: VisIt software users community <visit-users at elist.ornl.gov>
> Date: Monday, July 20, 2015 5:26 PM
> To: VisIt software users community <visit-users at elist.ornl.gov>
> Subject: Re: [visit-users] Launching Visit on Syrah
> Mark,
> I have been using the AMRvis program that is produced by the same group as the Boxlib.  It reads the plot files and displays the data from them.  It works fine on those files.  
> It seems to be extremely limited however in how you can view, manipulate, overlay, etc. the variables, which is why I was looking into Visit.  Which seems much more powerful than this program.
> Mark
> From: Miller, Mark C. [miller86 at llnl.gov]
> Sent: Monday, July 20, 2015 5:07 PM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
> Hi Mark,
> Do you by chance have any other "evidence" or reason to believe the boxlib file is "good." I mean, are there any other tools out their that digest boxlib and believe the file is ok? I'd just like a sanity check before I dig into this further. Obviously, the boxlib library being used in the VisIt plugin doesn't like the file for some reason.
> Mark
> From: <Converse>, "Mark C." <converse3 at llnl.gov>
> Reply-To: VisIt software users community <visit-users at elist.ornl.gov>
> Date: Monday, July 20, 2015 4:23 PM
> To: VisIt software users community <visit-users at elist.ornl.gov>
> Subject: Re: [visit-users] Launching Visit on Syrah
> Mark,
> That is a good question.  I'm using the latest version of Boxlib.  I just used git to grab it from the website a couple weeks ago.  In the documentation it says it is version  BoxLib Version 2014-02-28.  It's a little vague, but I assume probably not version 1?  Is there something we can do to update the plugin version?
> Mark
> From: Miller, Mark C. [miller86 at llnl.gov]
> Sent: Monday, July 20, 2015 11:28 AM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
> I've tried opening plt00100/Header on two different machines and get
> BoxLib::Error::Sorry, out of memory, bye ... !!!
> Eventually, the engine hits a SIGBRT (abort).
> Just prior to that, in the engine logs, I see this. . .
> avtDatabaseFactory: specifically told to use Boxlib3D_1.0
> Skipping already loaded database plugin Boxlib3D version 1.0
> Trying to open the file with the Boxlib3D file format, strict mode is off
> Would like to have issued warning "Reader does not currently support ghostzones.", but no callback was registered.
> signalhandler_core: SIGBRT!
> The file is only 25 Mb (compressed) so I assume Boxlib isn't really running out of memory and something else bad is happening.
> VisIt seems to know enough to think this is a 3D boxlib file and not 2D as I tried both readers and the 2D reader failed due to incorrect dimension.
> Has boxlib changed its file format recently? Are you maybe using a newer version of Boxlib than VisIt is using?
> Mark
> From: <Harrison>, "Cyrus D." <harrison37 at llnl.gov>
> Reply-To: VisIt software users community <visit-users at elist.ornl.gov>
> Date: Friday, July 17, 2015 2:51 PM
> To: VisIt software users community <visit-users at elist.ornl.gov>
> Subject: Re: [visit-users] Launching Visit on Syrah
> Hi Mark,
> I am sorry I haven’t had a chance to dive in — I’ll be out of the office next week, but I forwarded your data to the other LLNL VisIt developers to look at next week.
> -Cyrus
> On Jul 16, 2015, at 1:45 PM, Converse, Mark C. <converse3 at llnl.gov> wrote:
> Cyrus,
>   
> Any luck with opening that file format?
>   
> Thanks,
> Mark
>   
> From: Harrison, Cyrus D. [mailto:harrison37 at llnl.gov]
> Sent: Friday, July 10, 2015 4:32 PM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
>   
> Mark,
>   
> Thanks for sending me the file, this crashes on my mac as well — so I don’t think the issue is related to Syrah.
>   
> I’ll look into it further. 
>   
> -Cyrus
>   
> On Jul 10, 2015, at 10:31 AM, Converse, Mark C. <converse3 at llnl.gov> wrote:
> Cyrus,
>   
> Sure, how should I give it to you?  Email or give command on the LC machines?
>   
> Thanks,
> Mark
>   
> From: Harrison, Cyrus D. [mailto:harrison37 at llnl.gov]
> Sent: Thursday, July 09, 2015 8:21 PM
> To: VisIt software users community
> Subject: Re: [visit-users] Launching Visit on Syrah
>   
> Hi Mark,
>   
> Sorry, you are having so much trouble.
>   
> I don’t think we have a host-profile setup for syrah - this means it’s likely means you are running a serial engine.
> Depending on the size of your dataset, that could explain why you are running out of memory.
>   
> Would it be possible for you to give us part of your data set so we can investigate more?
>   
> -Cyrus
>   
> On Jul 8, 2015, at 1:45 PM, Converse, Mark C. <converse3 at llnl.gov> wrote:
> I’ve been trying to launch visit on syrah . llnl . gov.  I got it to open a metadata server on syrah and went to the folder where the data was.  Then I opened the file, and it crashed saying that it couldn’t find a compute engine and that would try to start a new one.  Then I get a
>   
> BoxLib::Error::Sorry, out of memory, bye…!!!!
>   
> Any idea how I can make this work.  According to the Boxlib website, their output is compatible with Visit.
>   
> Regards,
> Mark
> --
> VisIt Users Wiki: hxxp://visitusers.org/
> Frequently Asked Questions for VisIt: hxxp://visit.llnl.gov/FAQ.html
> To Unsubscribe: send a blank email to visit-users-unsubscribe at elist.ornl.gov
> More Options: hxxps://elist.ornl.gov/mailman/listinfo/visit-users
>   
> --
> VisIt Users Wiki: hxxp://visitusers.org/
> Frequently Asked Questions for VisIt: hxxp://visit.llnl.gov/FAQ.html
> To Unsubscribe: send a blank email to visit-users-unsubscribe at elist.ornl.gov
> More Options: hxxps://elist.ornl.gov/mailman/listinfo/visit-users
>   
> --
> VisIt Users Wiki: hxxp://visitusers.org/
> Frequently Asked Questions for VisIt: hxxp://visit.llnl.gov/FAQ.html
> To Unsubscribe: send a blank email to visit-users-unsubscribe at elist.ornl.gov
> More Options: hxxps://elist.ornl.gov/mailman/listinfo/visit-users
> --
> VisIt Users Wiki: hxxp://visitusers.org/
> Frequently Asked Questions for VisIt: hxxp://visit.llnl.gov/FAQ.html
> To Unsubscribe: send a blank email to visit-users-unsubscribe at elist.ornl.gov
> More Options: hxxps://elist.ornl.gov/mailman/listinfo/visit-users
>  
>  
> --
> VisIt Users Wiki: http://visitusers.org/
> Frequently Asked Questions for VisIt: http://visit.llnl.gov/FAQ.html
> To Unsubscribe: send a blank email to visit-users-unsubscribe at elist.ornl.gov
> More Options: https://elist.ornl.gov/mailman/listinfo/visit-users

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 3630 bytes
Desc: not available
URL: <https://elist.ornl.gov/pipermail/visit-users/attachments/20150731/20107e65/smime.p7s>


More information about the visit-users mailing list