Copyrighrt (c) 2014-2020 UPVI, LLC
Welcome, Guest
Username: Password: Remember me
LVHDF5 Toolkit Forum
  • Page:
  • 1
  • 2

TOPIC:

memory consumption 5 years 9 months ago #39

  • jdsommer
  • jdsommer's Avatar
  • Offline
  • Administrator
  • Administrator
  • Posts: 23
  • Thank you received: 1
Hi aluder,

Two more things to try:

1) You can explicitly force HDF5 to close out all of its objects with a call to H5close. This is on the Library palette. I haven't tried it, and I have some concerns that the next call to HDF5 following a call to H5close will fail. But, give it a shot.

2) If you're doing a logging-type operation where you are appending to a dataset, then you are doing dataset chunking. You may need to adjust the size of your chunks for better memory performance. See

www.hdfgroup.org/HDF5/faq/perfissues.html

This probably won't fix the problem, since you say that you see if in the Write Cluster Array example, but it could help. 

Jason

Please Log in or Create an account to join the conversation.

memory consumption 5 years 9 months ago #38

  • aluder
  • aluder's Avatar Topic Author
  • Visitor
  • Visitor
Dear Jason, Thank you for taking care of this. "periodically closing and reopening the dataset and file" is actually the first thing I have been playing with. In my case, closing and reopening didn't help. Since at the moment my only option is to completely close and restart LabVIEW, I wonder, if there's a "better" way to unload the hdf libraries?
Attachments:

Please Log in or Create an account to join the conversation.

Last edit: by aluder.

memory consumption 5 years 9 months ago #37

  • jdsommer
  • jdsommer's Avatar
  • Offline
  • Administrator
  • Administrator
  • Posts: 23
  • Thank you received: 1
Hello aluder,

Thanks for bringing this to my attention. I will attempt to replicate this, and if I can, then I'll try to track down the source of the problem. I certainly won't completely rule out the possibility of memory leaks in my code.

In the mean time, you might try periodically closing and reopening the dataset and file. Make sure to close all open references when you do this. It is conceivable the there is a leak associated with open HDF5 handles, and that closing them may deallocate the memory. If you try this, let me know if you notice any improvement.

Jason

Please Log in or Create an account to join the conversation.

memory consumption 5 years 9 months ago #36

  • aluder
  • aluder's Avatar Topic Author
  • Visitor
  • Visitor
Hi. My data write .vis all consume more and more memory. After a few days LabVIEW hangs due to several GB of memory consumption. When comparing my function with one of your examples, I found the same problem in the example: When putting "HDF5 Write Cluster Array Data.vi" in examples\UPVI\LVHDF5\basic into an endless loop, I can see an increase of memory usage in task manager. Can I do something about that?
I am using LV2014 fall with the HDF5 library 1.1.0.85

Please Log in or Create an account to join the conversation.

Last edit: by aluder.
  • Page:
  • 1
  • 2
Time to create page: 0.162 seconds