[msha]

During parallel run the term "ignSites.cells()" in file "ignite.H" returns no cell and ignition does not occur.
it does not depend on the number of cpus or the method of decomposition.
but with a single cpu there is no problem.

Any suggestions!!!

[mattijs]

Does it work in 1.5.x?

[msha]

IS OF 1.5.x modified for parallel?

I've not examined it,but I'll check it and inform you soon.

[msha]

YES!
Problem persists even running on OF 1.5.

[henry]

My understanding is this problem is fixed in version 1.5.x, have you tried that version?

H

[msha]

NO!
Would you tell me ,please;
1-what is the origin of such a problem?
2-what changes has been made in 1.5.x?

Thanks Henry!

[msha]

IS the problem solved in 1.5.x ?



would anyone tell me how to get tihs fix,I've "time out error in connection" when trying to use git.

thanks one who replies!

[msha]

I ran my cases on OF 1.5.x (the update up to the time of writing this)

The problem of ignition still persists.

I think that the function "findCell" in /src/meshTools/meshSearch/meshSearch.C doesn't work correctly in parallel (more than 2 CPUs)

the fix offered by mattjis (from git) in file "ignitionSites.C" was not helpful.

I hope this bug will be fixed soon by some one.

thanks!

[msha]

has any one fixed this up to this date?

[henry]

I have just run several XiFoam cases in serial and parallel and I do not see any problem: I get the correct number of ignition cells in serial and parallel.

H

[mattijs]

Does this problem occur with any of the tutorials or do you have a testcase we can try?

[msha]

Dear Henry & Mattijs:

My cases are :
1-Tutorial of XiFoam
2-Tutorial of dieselFoam
3-Two reactingFoam cases with complex burner geometry to simulate non-premixed combustion.

If you want,I can send my two reactingFoam cases as well as my customized solver(i.e. reactingFoam with ignition implementation), but the files are too big(one 500 Mb and another 50Mb )


Ignition is applied to reactingFoam solver as proposed by Niklas Nordin at http://openfoamwiki.net/index.php/Contrib_reactingFoam

*Notice that ignition fails in parallel run with more than 2 CPUs! I've tried all cases with 4,8,12,16,20 CPUs.


My cluster OS is CentOS with ROCKs as cluster tool.Codes like Fluent runs fine on the cluster.


There is some information on this in the threat:
http://www.cfd-online.com/OpenFOAM_D...tml?1229581879

Special Thanks for your time and consideration!
Sincerely
M.Sha

[msha]

Dear Henry & Mattijs:

My cases are :
1-Tutorial of XiFoam
2-Tutorial of dieselFoam
3-Two reactingFoam cases with complex burner geometry to simulate non-premixed combustion.

If you want,I can send my two reactingFoam cases as well as my customized solver(i.e. reactingFoam with ignition implementation), but the files are too big(one 500 Mb and another 50Mb )


Ignition is applied to reactingFoam solver as proposed by Niklas Nordin at http://openfoamwiki.net/index.php/Contrib_reactingFoam

*Notice that ignition fails in parallel run with more than 2 CPUs! I've tried all cases with 4,8,12,16,20 CPUs.


My cluster OS is CentOS with ROCKs as cluster tool.Codes like Fluent runs fine on the cluster.


There is some information on this in the threat:
http://www.cfd-online.com/OpenFOAM_D...tml?1229581879

Special Thanks for your time and consideration!
Sincerely
M.Sha

[henry]

I just ran the XiFoam tutorial on 4 CPUs without any issue with the ignition cells; I cannot reproduce your problem. If you would like us to investigate your problem further we will need more details.

H

[msha]

Dear Henry

What kind of details do you need to know?

[msha]

Dear Henry

My cluster is a AMD based cluster.each node has 4 CPUs(2 x Dual Core) 64 bit. The OS is CentOS and The cluster tool is ROCKS.

I've set WM_64=yes in $HOME/.bashrc to tell OF the system is 64.
As OF uses mpi I type for example:

mpirun -machinefile hosts -np 16 reactingFoam . boiler -parallel

Is above syntax true?

Then when code runs at first and before any time step it reports that Found Ignition Cells and writes cell number included in ignitionSites defined in combustionPropertiesDict.

But in time step it does not print the report "igniting cell (cell number)"
as when in single cpu run.

any comments?

P.S: My MS thesis depends on parallel processing with OF,So it would be very nice of you if you take my dilemma in your consideration.

Thanks for your Time!

[henry]

> mpirun -machinefile hosts -np 16 reactingFoam . boiler -parallel

> Is above syntax true?

For version 1.5 and 1.5.x this is not correct, see manual.

> But in time step it does not print the report "igniting cell (cell number)"

That's because the code is setup to print this information only from the master processor and if you do not have ignition cells on the master nothing will be printed but the code will run correctly nevertheless.

If you would like all the processors to print this message change "Info" to "Pout" in OpenFOAM-1.5.x/src/engine/ignition/ignite.H and recompile your application.

H

[msha]

Thanks Henry

I should tell you these earlier.

thanks again.