PostCohesive bugs for additional inter-particle forces when built with shared memory parallel

MFiX
,
#1

Hi everyone

I have edited a liquid bridge force and added it into calc_force_dem and calc_collision_wall_mod. By referring the original VDW cohesive force, the ratio of liquid bridge force and gravity acting on a particle is also assigned to PostCohesive, in order to check the magnitude of liquid bridge force in postprocessing.

image
image694×23 1.7 KB

image

The PostCohesive presents non-zero values after 0.01s if the shared memory parallel in the build solver is not selected.

image

image
image270×701 59.3 KB

However, the PostCohesive equals to zero and does not present nonzero values with iterations going on.

image

image
image266×700 65.2 KB
image
image268×706 61.1 KB

How to solve this problem? The attached zip file includes my code files and the mfix file.
test.zip (840.7 KB)

The above case is run on a Win10 system.

#2

It has been nearly two weeks and I have tested this case in a Linux system (CentOS7) with MFIX-21.1.4. The above problem disappears and reliable results can be get when DMP (available in Linux system) is used, while it still exists when built by SMP. Sometimes, the solver even stops running and reports the following bugs if SMP is used. I am puzzled about these phenomena although the solver can be built successfully with both SMP and DMP.

Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.
Backtrace for this error:
** Something went wrong while running addr2line. **
** Falling back to a simpler backtrace scheme. **
#28 0x7f49b185a2d2
#29 0x7f49b185aa0e
#30 0x7f4a04dcf3ff
#31 0x7f49b1f69f6e
#32 0x7f49b1f6bd42
#33 0x7f49b24e0961
#34 0x7f49b222fc39
#35 0x7f49b222fb23
#36
0x7f49b1f5f29d
#37 0x7f49b1f43602
#38 0x7f49b1f618dd
#39 0x558f6bd6588b
#40 0x558f6bd662d0
#41 0x558f6bdad601
#42 0x558f6bd1ed16
#43 0x558f6bd660c4
#44 0x558f6bda9380
#45 0x558f6bd1ed16
#46 0x558f6bd660c4
#47 0x558f6bda9380
#48 0x558f6bd000a5
#49 0x558f6bd15040
#50 0x558f6bd007b5
#51 0x558f6be07a22
#52 0x558f6bdc0a76
#53 0x7f4a0516eea4
#54 0x7f4a04e97b0c
#55 0xffffffffffffffff

Sometime, the solver reports the following bugs when running with DMP, but the solver can continue running.

image

image
image737×42 8.33 KB

#3

Please give steps for reproducing the SMP failure. How many threads are you running, etc. Please attach files if anything has changed.

Thanks,
– Charles

#4

Hi Charles, I am debugging programs and they will be uploaded here. Thanks for your help.

#5

Hi Charles

The bugs repoeted by the solver with SMP and my original case file are described and uploaded in this post.
The monitored velocity on one inlet planes are not equal to the given ones - MFiX - NETL Flow Science Software Support (doe.gov)

#6