[FLASH-USERS] larger FLASH hdf5 files in visit

[Sean Couch]

Hi Jacob,

I don’t have a solution, but I will chime in and say that you’re not crazy and not doing anything wrong. I’ve seen exactly this type of behavior with VisIt many times.  It’s like each process allocates enough memory for the entire dataset…  What’s more is that there are clearly memory leaks to boot.  For example, when making movies of a series of large FLASH plot files, I would see memory crashes after roughly the same number of frame creations.  Then restarting the movie making from where it left off would work fine, until the magic number of frames had again been made…  

As far as file size go, VisIt should have no problem with ~16 Gb files.  I routinely visualize plot files that are 100s of Gb in size.  But doing it in parallel is always a pain…  

I’ve looked briefly into the FLASH reader for VisIt and I don’t think the bug is there.  The reader is extremely simple.  I think the bug must be in the VisIt source itself.  And it’s been there for a very long time…

Sorry, this is not helpful but I don’t have a good solution.  Except to switch to yt as John suggests (but in my experience that’s no panacea, either…).

Cheers,
Sean


----------------------------------------------------------------------------------------------------------
Sean M. Couch
Assistant Professor
Department of Physics and Astronomy
Department of Computational Mathematics, Science, and Engineering
National Superconducting Cyclotron Laboratory/Facility for Rare Isotope Beams
Michigan State University
567 Wilson Rd, 3250 BPS
East Lansing, MI  48824
(517) 884-5035    ——    couch at pa.msu.edu <mailto:couch at pa.msu.edu>    ——    www.pa.msu.edu/~couch <http://www.pa.msu.edu/~couch>





> On Jul 8, 2016, at 10:02 AM, John ZuHone <jzuhone at space.mit.edu> wrote:
> 
> Hi Jacob,
> 
> I know this question is about VisIt, but I should point out that FLASH support for VisIt has lingered for quite some time now. 
> 
> yt is a pretty powerful package for dealing with FLASH data—it’s one of the data types that has the fullest support in yt, and you can do a lot more than you can in VisIt.
> 
> Find out more at http://yt-project.org <http://yt-project.org/>, or ping me off-list at john.zuhone at cfa.harvard.edu <mailto:john.zuhone at cfa.harvard.edu> if you have more questions. 
> 
> Best,
> 
> John Z
> 
>> On Jul 8, 2016, at 9:59 AM, McFarland, Jacob A. <mcfarlandja at missouri.edu <mailto:mcfarlandja at missouri.edu>> wrote:
>> 
>> All,
>> 
>> I just wanted to ping the community one more time to see if anyone has any suggestions on this issue. I have not received any feedback yet but I have received emails from some other users with this issue. 
>> 
>> Thanks for any help you can provide. 
>> 
>> - Jacob
>> From: flash-users-bounces at flash.uchicago.edu <mailto:flash-users-bounces at flash.uchicago.edu> [flash-users-bounces at flash.uchicago.edu <mailto:flash-users-bounces at flash.uchicago.edu>] on behalf of McFarland, Jacob A.
>> Sent: Tuesday, July 05, 2016 9:00 AM
>> To: flash-users at flash.uchicago.edu <mailto:flash-users at flash.uchicago.edu>
>> Subject: [FLASH-USERS] larger FLASH hdf5 files in visit
>> 
>> Hello all,
>> 
>> I emailed the visit users group with this issue but I have not received any info back yet. As this is a problem many of you may have encountered with FLASH data I thought the FLASH users group might have some advice as well. 
>> 
>> I am having a problem with Visit consuming a lot of memory when opening FLASH hdf5 files. I have simulations with about 50E6 nodes on the fluid mesh and about 20k particles. The hdf5 files are written with parallel hdf5 (I ran the setup with +hdf5TypeIO). The chk files are 5-16gb or so in size, getting larger in time as there are more nodes at later times. 
>> 
>> When I open the files with Visit each processor seems to consume enough memory to open the whole chk file no matter how many processor I throw at it. This means I end up memory limited and have to rune with very few processor on each computer node. 
>> 
>> I expected visit would be able to break up the hdf5 files so that each processor would not need the whole file in memory. This is the behavior I have seen when using other codes with data in the silo format. 
>> 
>> Is this the normal behavior for visit with large FLASH data files? 
>> 
>> If not, any suggestions on what I am doing wrong?
>> 
>> If so, any suggestions on how to mitigate this memory problem?
>> 
>> Thanks,
>> Jacob
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20160708/f7eae517/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1389 bytes
Desc: not available
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20160708/f7eae517/attachment.p7s>