[skinduptruk]

Originally Posted by MadsR

Hi Kurt.

I am glad the script worked out. I do not understand your git issue and why the r-directory should be added to the url :-) ?

With regards to the wiki I can't help, I guess everything should work when the script runs to the end without errors, no?

/Mads

Hi Mads,

nice to get a reply from the script author

the good news is that the allwmake worked, and the test was passed with systems ok.

i was able to run the cavity and dambreak tutorials as a test. can you recommend a "tutorial guide" to understand the others?

another question i have is that paraview help failed with a "failed to start qt assistant". does this mean my ubuntu 9.04 (latest) is missing a package or it is mis-configured? sorry i have not researched more myself - does it work for you?

[MadsR]

Hi Kurt.

I can't really recommend a tutorial guide as I aren't aware know of such. But what is the problem you are trying to solve? Maybe the good readers of this forum can point you in the right direction - but maybe it's worth a new thread, since it's a bit off topic.

Paraview "should work" as everything else should ;-) but you might miss a QT library, which is odd, as the script is installing QT. Could you provide the full output from paraview (if more) and also tell us if paraview is failing to run, or just is spitting out some errors when running?

/Mads

[airfoil]

Hi Mads,
Thanks for the script. It is such a good script. However I have to change the git directory from git clone git://repo.or.cz/OpenFOAM-1.6.x.git to
git clone http://repo.or.cz/r/OpenFOAM-1.6.x.git like mentioned by Kurt. Actually both directory are given by OpenCFD in it website. Somehow I have problem with the first entry. After I change the git directory, the script work without any flaw for me.

Rgds,
Airfoil

[awallin]

any chance that the script could give some progress-bar or report on how the compile is progressing? Now I'm just staring at "THIS CAN TAKE HOURS" and don't know what's going on...

are there parts of the make process that could be done in parallel? now I have 7 cores doing nothing while I stare at the above text...

otherwise this seems to work on a fresh Ubuntu 9.04 install.
Is there a detectable performance difference between the binary version from sourceforge and this one pulled from the git repository and compiled on the hardware it will run?

AW

[akidess]

Anders, I think cmake puts the progress in square brackets, so if you tail the log file, you should see stuff like [14%] ... Anything past that is very non-trivial.

About the performance difference I just have this link: http://www.afs.enea.it/funel/CFD/Ope...amcomptest.htm

[akidess]

Ok, I modified the 64bit version of the script to enable parallel compilation. This modification is completely untested, so I'd be thankful if someone would be willing to try it out.

Download: installOF64

PS. I forgot to mention - to compile in parallel, call the script with the number of cores as argument, e.g. "./installOF64 4"

[awallin]

Originally Posted by akidess

Ok, I modified the 64bit version of the script to enable parallel compilation. This modification is completely untested, so I'd be thankful if someone would be willing to try it out.
Download: installOF64

on my machine the number of cores ends up in $1, so I had to change that to make it work. also I think there's a space missing inside the right bracket in the if statement:
if [ $1 -gt 1 ]

with these modifications I do seem to get four wmakeScheduler processes running, and some output to make.log
but after a while it seems all the processes sleep and nothing more gets added to make.log (or am I just impatient and should let it run for hours?)
after nothing happens (0% load on all cores for many minutes) I interrupt with ctrl-c and see this in make.log:
make: *** [Make/linux64GccDPOpt/bool.o] Interrupt
make: *** [Make/linux64GccDPOpt/clock.o] Interrupt
make: *** [Make/linux64GccDPOpt/argList.o] Interrupt
make: *** [Make/linux64GccDPOpt/dimensionedConstants.o] Interrupt

[akidess]

Thanks, I corrected the script (forgot to start counting from 0!) and tested the compilation part on one of my folders in OpenFOAM-1.5. It seems to work for me, using four cores the compilation definitely gets done in shorter time.

[MadsR]

@anton: BRILLIANT. I'd include your parallel execution bit.

@anders: good idea with the progress bar, maybe we can grep some of the output out to the screen to inform the user on what's going on. I will look into that.

/Mads

[menorka]

I'm trying out the parallel script as I type. The CPU utilisation is a bit on the low side - around 10% on two cores, compared to 100% on one core with the old script. Looks fishy.

[awallin]

Originally Posted by menorka

I'm trying out the parallel script as I type. The CPU utilisation is a bit on the low side - around 10% on two cores, compared to 100% on one core with the old script. Looks fishy.

did you ever get it to compile? on my machine it seems all the scheduler processes just sleep and nothing happens.

[menorka]

I canceled it and am currently compiling with a single core.

[MadsR]

Hi.

I made the scripts back to single core, as safe install wins over speed.

/Mads

[akidess]

Originally Posted by MadsR

Hi.

I made the scripts back to single core, as safe install wins over speed.

/Mads

Yes I agree, especially since it's a one time thing. Still a bit odd it's working for me but not for others. I wonder if it's because I tested it with 1.5...

[akidess]

Oh, and about the progress: Looking at the make.log there is nothing really that would indicate how far we are with compiling.

[MadsR]

Anton, yeah it's weird. I have modified the script multiple times since I published it, eventhough its initial version installed perfectly on my Ubuntu - some installations out there are just different somehow.

/Mads

[menorka]

Can't get paraFoam to work despite running the old script.

On Ubuntu 9.10...

• Ran installOF64. foamInstallationTest reported everything OK.
• mkdir -p $FOAM_RUN
• cp -r $WM_PROJECT_DIR/tutorials $FOAM_RUN
• cd $FOAM_RUN/tutorials/incompressible/icoFoam/cavity
• blockMesh
• icoFoam

Everything works until...

• paraFoam

paraView starts up but I get a selection box "Open Data With". Checking make.log I find this as the last entry...

make[1]: Leaving directory `/home/linus/OpenFOAM/OpenFOAM-1.6.x/applications/utilities/thermophysical'
make: Target `application' not remade because of errors.

If I keep looking through the file I also find this...

cmake: /home/linus/OpenFOAM/ThirdParty-1.6.x/gcc-4.3.3/platforms/linux64/lib64/libstdc++.so.6: version `GLIBCXX_3.4.11' not found (required by cmake)
+ make
make[3]: Entering directory `/home/linus/OpenFOAM/OpenFOAM-1.6.x/applications/utilities/postProcessing/graphics/PV3FoamReader/PV3FoamReader/Make/linux64GccDPOpt'
make[3]: *** No targets specified and no makefile found.  Stop.
make[3]: Leaving directory `/home/linus/OpenFOAM/OpenFOAM-1.6.x/applications/utilities/postProcessing/graphics/PV3FoamReader/PV3FoamReader/Make/linux64GccDPOpt'
make[2]: *** [PV3FoamReader] Error 2

and this...

make[3]: Leaving directory `/home/linus/OpenFOAM/OpenFOAM-1.6.x/applications/utilities/postProcessing/graphics/ensightFoamReader'
make[2]: Target `application' not remade because of errors.
make[2]: Leaving directory `/home/linus/OpenFOAM/OpenFOAM-1.6.x/applications/utilities/postProcessing/graphics'
make[1]: *** [graphics] Error 2
make[2]: Entering directory `/home/linus/OpenFOAM/OpenFOAM-1.6.x/applications/utilities/postProcessing/lagrangian'

Trying to rerun makeParaView doesn't work. Any and all suggestions are welcome.

[akidess]

I wish people would use the LTS versions more, but then again I can't complain since I'm using 8.10 myself. It looks like this is an issue because for 9.10, ubuntu has switched from gcc 4.3 to 4.4 (see also http://www.cfd-online.com/Forums/ope...not-found.html). I think you'll have to edit your configuration to use the system compiler and recompile, or rebuild cmake with gcc 4.3.

[MadsR]

Uh. There is a new version of Ubuntu. They never stop, do they??? ;-)

Anton, what do we do? You seem to know about Ubuntu releases, LTS and stuff. Is the script not working anymore in 9.10 (as menorka reports), or is something else wrong?


/Mads

[akidess]

To me it looks like the script does not work any more on 9.10. It's hard to fix though without having 9.10 myself.

I'm assuming that the system cmake is being used (menorka: post "which cmake"). If so, a solution would be to edit $PATH so that the OpenFOAM cmake is used. Judging from other posts in the forum that might be problematic too, and cmake would have to be built from source.

For now new users might want to hold back from using 9.10 until an experienced user can resolve the problem. I might try figuring it out myself in a VM if I find the time.