Issues with latest release (6.7.8) - SPEC variables as nitrogen and very slow simulations

410 views
Skip to first unread message

John S.

unread,
May 27, 2022, 9:11:40 AM5/27/22
to FDS and Smokeview Discussions
Hi all,

I am having two issues with running codes that ran fine on 6.7.7.

1. First is this warning:

WARNING: SPEC PARAFFIN VAPOR is not in the table of pre-defined species. Any unassigned SPEC variables in the input were assigned the properties of nitrogen.

I get this for a species that is defined as follows:

&SPEC ID = 'PARAFFIN VAPOR'
      FORMULA               ='C21H27N1O3'
      LUMPED_COMPONENT_ONLY = .TRUE./


This species is used as REAC FUEL as a lumped component of a 7 component mixture.  Can I ignore this warning?

2. The second issue is that a code that previously ran on 6.7.7 on the same computer in about 5 days now looks like it will be too slow to ever finish, just creeping by the hundredth of a second.  It is an 18 mesh code called by 18 MPI processes.  I do not have hyperthreading disabled on this computer.  Is that the source of my issue?

Thanks for any help!

Glenn Forney

unread,
May 27, 2022, 9:30:06 AM5/27/22
to FDS and Smokeview Discussions
probably not the issue but an issue.  we always disable hyper threading on computer we use to run fds.  I would disable it .  I would be surprised if fds speed up by more than 2x though after you did this

dr_jfloyd

unread,
May 27, 2022, 9:32:12 AM5/27/22
to FDS and Smokeview Discussions
In your case, you should not ignore that warning. Please read the user's guide discussion on the default method via which FDS defines the specific heat of a user defined species when no user defined specific heat is provided.

John S.

unread,
May 27, 2022, 10:06:55 AM5/27/22
to FDS and Smokeview Discussions
I will try disabling hyperthreading and see if I made any mistakes anywhere else.  Let you know.

I will also define Specific Heats for all my REAC FUEL SPEC IDs.

Thanks for the help!

John S.

unread,
May 27, 2022, 10:47:31 AM5/27/22
to FDS and Smokeview Discussions
I tried disabling hyperthreading, and I think it actually got slower.  I tried a code that ran before with a larger mesh size and similar results.  Chugging away by .01s simulation time.

Any suggestions?

Thanks!

John S.

unread,
May 27, 2022, 10:50:15 AM5/27/22
to FDS and Smokeview Discussions
I mean, very slow, like reaching the 1000 timestamp at simulation time  .19 seconds, effectively never going to finish.  This same code took 7 hours on 6.7.7, everything else identical.

John S.

unread,
May 27, 2022, 10:58:33 AM5/27/22
to FDS and Smokeview Discussions
I noticed when it installed there was an error that is could not create Desktop shortcut CMDfds.lnk.  Does this sound familiar?

Glenn Forney

unread,
May 27, 2022, 6:02:59 PM5/27/22
to FDS and Smokeview Discussions
try this
1.  open up a command shell and type 
where cmd
on my computer it says C:\Windows\System32\cmd.exe
2.  with windows explorer go to the directory containing cmd.exe (ie open the C: drive , then the Windows folder the the System32 folder)
3.  right click on the file cmd.exe and select shortcut 
4. it will say it cant create the shorcut would you like to put it on the desktop .  say yes
5. go to the desktop and right click on the shortcut you just created and select properties
6.  in the "Target" section you sould see C:\Windows\System32\cmd.exe
add /k fdsinit at the end
and click on Appy or Ok

this is the same shortcut that the fds installer should have put on your desktop.

plan B
1.  open up a regular CMD shell
2. type fdsinit
this is exactly what the CMDfds shortcut does

I don't know why the installer is able to put the CMDfds shortcut on some computers and not others.

John S.

unread,
Jun 1, 2022, 8:45:44 AM6/1/22
to FDS and Smokeview Discussions
Thanks!  It looks like you can also just type CMDfds in the search bar and the shortcut is the first result.


About fds MPI process running extremely slow on 6.7.8.:  I looked through documentation section 3.2 and some of the history of FDS Discussion Group.  I am thinking either my mesh code is written in an unfriendly way or I need to use a scheduler?  I am running Windows 10 and the code worked fine in 6.7.7.  Please see my mesh here:

&MESH IJK=15,15,26, XB=-.046,-0.016,-.046,-0.016,0,.052, MULT_ID='mesh' /
&MULT ID='mesh', DX=0.03, DY=0.03, DZ=0.052, I_UPPER=2, J_UPPER=2, K_UPPER=1 /


I saw some other users were assigning each mesh as a line and assigning each processor to a mesh in the .fds file itself.  Is this a best practice for using MPI on multiple meshes?

I also saw it was recommended to use a scheduler to run MPI on Windows with hyperthreading disabled.  Can you recommend if I need to try and do this, and what scheduler is best?

Thank you for any help!
Reply all
Reply to author
Forward
0 new messages