This problem occurs during the construction and solution process. It cannot be solved. Seek help from professionals
Error from C:/repo/mingw-w64-crt-git/src/mingw-w64/mingw-w64-crt/math/lround.c:36
Error: Solver crash!
The MFiX solver has terminated unexpectedly
Error information:
float invalid operation in lround
at C:/repo/mingw-w64-crt-git/src/mingw-w64/mingw-w64-crt/math/lround.c:36
This message about “Cannot find file” is not the source of the error, but it is a result of the GUI trying to print a stack trace after a solver crash. We catch FPE errors (floating point exceptions) and open the source file in the MFiX editor. Due to limitations on the Windows platform, we are not able to obtain a full stack but only the innermost function call, which, in this case, is happening inside a library function lround. Since this code is not part of MFiX, the source cannot be displayed, leading to the message you are seeing. What we need to know is what MFiX code called lround but this is not accessible on Window at this time (it is a long-term goal of mine to fix this but it’s not easy)
You can try turning off FPE traps in the Numerics/Advanced pane but this wlll just be hiding the problem
I’ll try running the case on Linux and see if I can get better debug info.
@JZH Did you build a custom solver for this case? In your uploaded ZIP file the filname is USR_RATES .f with a space before the .f - this will not work - the file should be called usr_rates.f
I see you are running with 12 SMP threads, did you try running the case in serial mode?
How long did you have to run before the error occurred? What was the simulation time?
I’m running with a custom solver (built with your code) on Linux with 8 threads, I haven’t seen the error yet, but the estimated remaining time is over 24 hours. I’ll let it keep running and see what happens.
I was able to run this job on Linux to completion (t=0.01s), in about 20 hours of running. On Windows I saw the error after about a minute (t=1e-7s).
Due to limited debugging tools on the Windows platform, I was not able to trace down the source of the error. I next tried building the solver in DEBUG mode, which also did not help, and running a serial rather than SMP case. Without SMP, the same FPE occurred.
I am sorry to say that until we get better ability to get full stack traces on Windows, I am not able to offer further help here. My only advice at this point is to try to run on Linux if posslble. You will have better debuggability, and your jobs will run faster as well. DMP is also available for Linux and often gives better performance than SMP.
