Fluids

Fluids

Unsorted Chunks Error

    • computationalphysics
      Subscriber

      Dear Team, we are seeing the error below while running an urgent 130 million cell (mixed poly and hex) model on an 18-node x 120 core (each) HB120RSV2 Linux cluster on Azure.  We are using 2020R1, automatic Metis partitioning, single precision, transient, linear eddy-viscosity based turbulent flow.  We have DPM engaged for future tracking, but no tracking is occurring now. I consider this a relatively simple physics embodiment with no underlying fundamental challenges. 


       


      *** Error in `/shared/ansys_inc/v201/fluent/fluent20.1.0/lnamd64/3d_host/fluent.20.1.0': free(): corrupted unsorted chunks: 0x0000000006eb95d0 ***


       


      A search on this forum revealed nothing substantial.  Someone is suggesting a lack of memory; however, memory should not be a problem on this many nodes.  


       


      A brief google search revealed:  


      :  “This error occurs because you have corrupted a pointer the system uses to keep track of space that has been malloc()ed. The most common causes for this type of corruption are (1) using an uninitialized pointer and (2) writing more data into memory than was allocated for the buffer into which the data is being written. “


      This appears to be a code problem and not a hardware problem.  Please help asap!

    • Karthik R
      Administrator

      Hello,


      Are you using any external UDF code in your simulation?


      Also, is there a reason for single precision? 


      Do you get this error at the beginning of your run or in the middle?


      Thank you.


      Karthik

    • computationalphysics
      Subscriber

      Dear Karthik,


      Thank you very much for your reply!  We are not using a UDF, though we will be eventually.  We are using SP, because there is no numerics- or physics-based reason to use extra RAM and solver speed reduction associated with DP.  This error occurs at seemingly "random" times in the simulation, and it, of course, kills the solver's parallel processes.  


      -Wayne

    • Karthik R
      Administrator

      Hello Wayne,


      Are you able to run a smaller model without any issues or are you seeing this happen on a smaller model as well? Also, I'm assuming that the model was built native to 2020R1. Is this correct?


      I did some digging into this error and it has historically been associated with either incompatible hardware or backward compatibility of the code. This is the reason for my questions. Please let me know.


      Thank you.


      Karthik


       

    • computationalphysics
      Subscriber

      Dear Karthik,


      Thank you for your continued dialogue.  Typically, my models are not built native to 2020R1, but this one is.  I will work on the smaller model test.  That is an excellent idea.


      -Wayne

Viewing 4 reply threads
  • You must be logged in to reply to this topic.