blueCFD / Core

Coordination repository for the blueCFD-Core: Issue tracking, Wiki, project webpage and miscellaneous scripts
http://bluecfd.github.io/Core
53 stars 10 forks source link

compressibleVoF/depthCharge2D tutorial case not working with blueCFD-Core 2024-1 #256

Open wyldckat opened 2 months ago

wyldckat commented 2 months ago

The tutorial case compressibleVoF/depthCharge2D from OpenFOAM 12 is failing on our port of blueCFD-Core 2024-1. Details:

  1. It crashes at Time = 0.427442735s

  2. Diverging time stamp:

    • Windows: Time = 0.3106666667s
    • Linux: Time = 0.310625s
  3. Looking at the results, it seems to be because there are certain regions that drop temperature to 140-180K, while the initial temperature was around 300K...

  4. Possibilities:

    a. It almost seems to be some memory garbage coming in from the top boundary; b. Or there is some threshold behavior that is not working the same way on Windows as it does on Linux; c. Or there is a variable that stays undefined during said threshold calculation.

wyldckat commented 2 months ago

Further tests conducted so far

  1. Using MSYS2 from blueCFD-Core 2020 to build OpenFOAM 12, namely using GCC 10.2.0, gives the same exact error.

  2. Tried compiling GCC 12.2.0 with the same stack as we are using on MSYS2, namely:

    cd $WM_THIRD_PARTY_DIR
    wget "https://raw.github.com/wyldckat/scripts4OpenFOAM3rdParty/master/getGcc"
    wget "https://raw.github.com/wyldckat/ThirdParty-2.0.x/binutils/makeBinutils"
    wget "https://raw.github.com/wyldckat/ThirdParty-2.0.x/binutils/getBinutils"
    chmod +x get* make*
    
    ./getGcc gcc-12.2.0 gmp-6.2.1 mpfr-4.2.0 mpc-1.3.1
    ./makeGcc gcc-12.2.0 gmp-6.2.1 mpfr-4.2.0 mpc-1.3.1 -no-multilib > log.mkgcc 2>&1
    wmRefresh
    
    ./getBinutils binutils-2.39
    ./makeBinutils gcc-12.2.0 binutils-2.39 > log.mkbinutils 2>&1
  3. The results from running with GCC 8.5.0 on RockyLinux 8.6 and GCC 12.2.0 (with -O2 -no-tree-vectorize) were identical.

  4. Tested with GCC 12.2.0 and FOAM_NAN=1 and still identical, although took slightly longer to run (200->215s).

wyldckat commented 1 week ago

Ideas and possibilities

  1. TODO Perhaps the fix made here: https://bugs.openfoam.org/view.php?id=4167 - will help solve this issue.